Who Does Your AI Agent Work For?

The Banks Just Answered, and Customers Lost

Who Does Your AI Agent Work For?

by Braden Kelley and Art Inteligencia


Picture a customer who is tired of managing her money. She hires an AI agent to watch her accounts, dodge fees and move cash to wherever it earns the most. She hands over her login, goes to dinner, and comes home to a drained account.

Whose problem is that? This week, a growing list of banks answered the question, and the answer was: hers.

I think this is one of the most important customer experience stories of the year, and it’s hiding inside a quiet update to some legal terms. It’s about who bears the risk when software acts on a customer’s behalf. But it’s also about something bigger, and more uncomfortable for every company that sells to people: your customer is about to stop being just a person.

The short version: Citi, PNC and Citizens updated their customer terms this week, joining JPMorgan and Bank of America, to say customers are responsible for what third-party AI agents do with their accounts, even if the agent is hacked. Citi’s own agents are carved out. The law underneath is unsettled, the banks’ reasons are real, and the experience design challenge is wide open.

What the banks actually did

According to TechCrunch, Citi updated its online user agreement this week so that customers who give third-party AI tools access to their accounts are responsible for activity through them. Citi’s terms also cover security failures: if the tool leaks credentials or is hacked, the customer is responsible for what the hackers do. PNC changed its terms the day of the report, Citizens emailed customers a similar warning, and JPMorgan and Bank of America announced similar policies last month. The banks say they won’t reverse transactions initiated by automated assistants, whether those were mistakes or not.

Here’s the detail I can’t stop thinking about: Citi’s own AI agents can be used without the customer taking on that added liability. TechCrunch doesn’t say whether the other banks offer a similar exemption. So, at least at Citi, the same task, performed by software, carries different risk depending on whose software it is.

Citi’s spokesperson said the bank wants customers “to understand the risks associated with sharing banking credentials.” That’s fair, as far as it goes.

Why the banks aren’t being irrational

I want to be fair to the banks, because their concerns aren’t imaginary.

  • The security stories are real. American Banker reports that at least five large banks (JPMorgan Chase, Citi, Bank of America, Wells Fargo and Morgan Stanley) are blocking external AI agents from customer accounts. It describes one unnamed bank whose monitoring caught an agent resetting a password, accessing an account, changing the approved transfer destinations and moving $100,000. That bank’s fraud team then recommended blocking all external agents.
  • Banks often can’t tell who’s acting. One security write-up, based on reporting from Axios and Reuters, put the problem in a line I like: “a valid login no longer proves that a person chose the action.” It reports no public breach of the named banks’ core systems and no confirmed customer losses so far, so this is risk management, not a response to a wave of victims.
  • There may be a competitive motive too. TechCrunch notes that banks are also worried AI tools could help customers avoid fees or low-yield deposit accounts. I can’t tell you how much that weighs against the security concerns, and the banks haven’t said. But it’s worth noticing that the people who profit from customer inertia are now facing a tool built to defeat it.

The law isn’t settled, and that’s the real story

I’m not a lawyer, and nothing here is legal advice. But the sources I read agree on one thing: nobody knows how this will shake out.

An October column in Payments Dive by financial services consultant Aaron Cheiffetz says it plainly: “There’s no clear legal rule for who bears the loss when an AI tool exceeds its mandate.” He notes that, under Regulation E, a consumer can be liable for transfers made by someone they gave access to, even if that person exceeded their authority, but that it’s unclear whether sharing credentials with an AI tool triggers that rule.

The law firm Goodwin reads the current rules in a way that helps the banks. In its analysis of agentic payments, “a transfer is presumed authorized if a consumer furnishes credentials to authorize an agent to make payments,” which means an agent that goes beyond what you intended can still produce a transfer the law treats as authorized. For credit cards, the standard is looser: a payment is authorized if the person initiating it has “actual, implied, or apparent authority.” Goodwin also notes that outcomes “depend on the facts: what the user said, what the agent understood, and what records exist.”

That last phrase matters, and I’ll come back to it. There’s also a policy gap underneath. The security write-up I mentioned says no federal rule defines delegated agent access, and that the CFPB’s open banking rule is enjoined. It also points out that in the U.K., most victims of push-payment fraud have been entitled to reimbursement up to £85,000 since October 2024. That’s the write-up’s characterization, so check it before you quote it. The point is that the U.S. is choosing a very different default.

The banks have a defensible position under today’s ambiguity. Contract terms are an easy tool to reach for when the rules are unclear. But “defensible” and “good customer experience” are different measures, and companies that confuse them tend to learn the difference the hard way.

The experience design problem nobody has solved

Here’s where I stop being a spectator and start being a designer.

Today’s banking experience, like most customer experiences, assumes a human at the keyboard. Login, authentication, fraud detection, error resolution, dispute processes: all of it was designed around a person who clicks a button, feels a certain way and can be called on the phone. Now a piece of software sits between the customer and the company. Both sides are improvising.

The customer’s real need hasn’t changed: do this for me, safely, within limits I set, and let me undo it if it goes wrong. What’s missing is the experience layer that makes that possible. Goodwin’s recommendations read like a design brief:

  • Define delegated authority precisely: amounts, merchants, categories, timing and how much discretion the agent has.
  • Keep retrievable records of what the customer reviewed and approved. A voice prompt or a checkbox alone isn’t enough.
  • Add pause-and-confirm checkpoints for unusual merchants, atypical amounts or new payment rails.
  • Let customers narrow or revoke authority, with changes taking effect before the next transaction.
  • Show the terms up front, including refund and return policies, before the agent completes a purchase.

Notice what that list isn’t. It isn’t “tell customers they’re on their own.” It’s a set of designed experiences: permission slips with limits, receipts, confirmation moments, an off switch. The banks that build those will be able to say yes to agents. The banks that don’t will be left saying no, while customers take their business, and their agents, somewhere less restrictive. As consultant Jeff McMillan told American Banker, if banks block agents, “the agents will go and open accounts for consumers at other firms that are not as restrictive.” He added: “Agents aren’t going anywhere.” American Banker also reports that some bankers think the block is only temporary.

The second problem: whose side is the agent on?

Liability is the headline, but there’s a quieter issue that applies to every company, not just banks. If an agent is choosing for your customer, what is it optimizing?

A Columbia and Yale research team studied this in a paper titled “What Is Your AI Agent Buying?” The authors built a sandbox where AI shopping agents chose among products and then tested how the agents behaved. Their findings should make any brand manager nervous:

  • The model matters enormously. The same fitness watch captured about 45% market share under one model, about 77% under a newer version of that same model, and only about 6% under a different provider’s model. The authors write that “model updates can drastically reshuffle market shares.”
  • Presentation quirks decide outcomes. Listing position had large effects that varied by provider and version. The agents consistently penalized “Sponsored” tags and rewarded “Overall Pick” endorsements.
  • Sellers can game it. AI-edited product descriptions raised market share in most of the tested buyer agents, with gains of up to about 15 percentage points.

Their conclusion: “agentic markets are volatile and fundamentally different from human-centric commerce.” Remember that these were lab tests, run on specific models in a sandbox, not live shopping. But the direction is clear. If your customer’s agent is the first thing evaluating your offer, then the agent’s quirks become part of your customer experience, and a model update you don’t control can change your sales overnight.

What this means for the rest of us

You don’t have to be a bank to be affected. If you sell, book, ship, support or bill, agents are coming to your front door. The questions that follow are experience questions:

Moment The question to ask
Identity Can we tell the difference between a customer, a customer’s authorized agent and a rogue one? What do we do when we can’t?
Authority Can customers set limits for an agent (amounts, categories, time) and see them recorded?
Confirmation Where does a human get pulled back in? Do we make that moment easy or irritating?
Error and recovery When an agent makes a mistake, what does our dispute and refund process look like? Who pays?
Discovery How do we show up when an agent, not a person, is doing the comparing? Is our information accurate and complete?
Fairness Do our terms treat customers’ agents differently from our own? Can we defend that to a customer’s face?
Trust If we’re building our own agents, can customers tell whose interests they serve?

These are exactly the kinds of gaps an experience audit is built to find: the places where the journey you designed meets the journey customers actually take. The next version of that journey includes a robot, and almost nobody has walked it yet.

Opportunities for innovators

Where there’s friction, there’s room to build. Here’s where I’d look:

  1. A standard “mandate” for agents. A portable, machine-readable record of what a customer authorized, which any bank or merchant can read and honor. Payments Dive’s author recommends exactly this kind of recorded limit.
  2. Shared-liability models. Cheiffetz argues losses should be allocated among issuers, networks, merchants and AI-tool providers without shifting the burden onto consumers. Whoever designs a fair, insurable version of that wins a lot of trust.
  3. Agent identity and kill switches. Ways to verify an agent, limit what it can do and stop it instantly.
  4. Verifiable loyalty. A way for a customer to know that their agent is working for them, not for whoever pays it a fee.
  5. Agent-ready experience design. Companies that make their information, policies and recovery paths clear to both people and machines will be the easy ones to do business with.

My verdict

The banks’ move is understandable, and I wouldn’t call it villainy. When the rules are unclear and the risks are new, protecting yourself in the contract is the reflex. But it’s a reflex, not a strategy.

The better question for any leader is: what would it take for us to say yes? Not “yes to anything,” but yes with limits, records, checkpoints and a fair answer to “who pays when it goes wrong?” The first organization in each industry to build that experience will turn a liability headache into a reason for customers to choose them.

Because here’s the thing I keep coming back to. The customer who sends a robot to the bank isn’t trying to be reckless. She’s trying to get her time back. A company that makes that safe is giving her something she values. A company that says “you’re on your own” is telling her what it thinks of her.

A note on what I could and couldn’t verify

The bank policy details come from news reports (principally TechCrunch), not from reading each bank’s agreement, so check the actual terms before quoting their wording. The JPMorgan and Bank of America details come mainly from secondary reporting. The legal commentary reflects individual authors’ readings of unsettled law, and none of it is legal advice. The Columbia and Yale findings come from a research sandbox, and its authors say the results vary by model and version.

Want to know what your experience looks like to an agent?

If you’d like an independent look at how your customer, employee and partner journeys hold up as AI agents join the process, that’s what the Customer Experience Audit is for. If you want to put a number on what friction costs you first, the CX ROI Calculator is free. And if you’d like to talk it through, get in touch.

Sources

Image Credits: Gemini

Content Authenticity Statement: The topic area, key elements to focus on, etc. were decisions made by Braden Kelley, with a little help from Claude to clean up the article.

Subscribe to Human-Centered Change & Innovation WeeklySign up here to get Human-Centered Change & Innovation Weekly delivered to your inbox every week.

The Best Way to Change Behavior

The Best Way to Change Behavior

GUEST POST from Mike Shipulski

There’s one good way to change behavior. But don’t take my word for it, take Daniel Kahneman’s, psychologist who was awarded the Nobel Prize in Economic Sciences. In Freakonomics Radio’s podcast How to Launch a Behavioral Change Revolution, Kahneman explains how to achieve change in behavior. His explanation is short [30:35 – 35:21] and good.

Kahneman describes a theory of Kurt Lewin, his academic grandfather, where behavior is an equilibrium, a balance between driving forces that push for change and restraining forces that hold back change. Kahneman goes on to describe Lewin’s insight. “Lewin’s insight was that if you want to achieve change in behavior, there is one good way to do it and one bad way. The good way is by diminishing restraining forces, not by increasing the driving forces. That turns out to be profoundly non-intuitive.”

Usually, when we want someone to change, we push them in the direction we want them to go. Kahneman says this approach is natural, but ineffective. He offers a different approach – “Instead of asking how can I get him or her to do it, it starts with a question of why isn’t she doing it already? Go one-by-one, systematically, and you ask ‘What can I do to make it easier for that person to move?’”

What would happen if instead of pushing someone to change, you understand what’s in the way and eliminated the restraining force? I don’t know, but I’m going to give it a try.

Kahneman goes on to describe how to make things easier for a person to move. He says “…the way to make things easier is almost always by controlling the individual’s environment…by just making it easier.” Sounds pretty simple – change people’s environment to make it easier for them to move toward the desired behavior. But, we don’t do it that way.

Kahneman gives more detail. “Are there incentives that work against it? Let’s change the incentives.” And then he gives a simple example. “I want to influence B, but there is A in the background and it’s A who is a restraining force on B, let’s work on A, not on B.”

I urge you to listen to the short segment to hear Kahneman’s words for yourself. His ideas really hit home when you hear them from him.

Image credits: Daniel Kahneman

Subscribe to Human-Centered Change & Innovation WeeklySign up here to join 17,000+ leaders getting Human-Centered Change & Innovation Weekly delivered to their inbox every week.

The Future of Customer Experience

AI First, Not AI Only

The Future of Customer Experience

GUEST POST from Shep Hyken

How can you not enjoy working with a company whose name is NiCE? Seriously, I’ve had the good fortune of working with NiCE for about 10 years. It is one of the largest AI companies for customer experience, working with businesses of all sizes, from SMBs to global enterprises. Recently, I’ve been working with Michele Carlson, senior product marketing manager, on several virtual events. Our most recent one was on the Top AI Trends CX Leaders Must Act On in 2026.

I loved the event so much that I did a follow-up interview to talk about a few of my favorite trends that would apply to most of our subscribers.

AI is shaping customer experience, yet people still love talking to live agents for customer support. As a matter of fact, my annual customer service and experience research finds that, given the choice between calling a company on the phone or using its digital options, including AI, 68% of American consumers prefer the phone. However, 47% of consumers said that even with their preference for the phone, they still want the digital option, and if it wasn’t provided, they would consider switching to a company that offered it. With that as a backdrop, here are three of my favorite trends from our conversation:

  1. AI-First CX Becomes the Dominant Model for Customer Engagement: This means we may suggest to our customers to use an AI solution, especially for simple tasks like checking on an order, making a change of address and other easy-to-do self-service tasks, but there should always, and I’ll emphasize the word always, be a quick and effortless way to connect to a live agent. Another way of saying this is: AI First, but Not AI Only.
  2. Agentic AI and LAMs (Language Action Models): AI is no longer just about conversation. Agentic AI can take action. For example, if I’m interacting with a chatbot about a problem, it can determine that I need a replacement and then perform the multiple time-consuming steps that humans normally do. AI updates the order, confirms shipping details, triggers warehouse fulfillment, generates necessary documents, and tracks the shipping, all without human intervention. This frees up employees to be more productive and efficient.
  3. Experience Memory Compounds AI Value: This is your secret weapon to creating loyalty with your customers. AI allows you to know your customers better than ever. It remembers everything, tracking sales history, buying patterns, preferences, and more. Chatbots can personalize the experience, almost as if they were human. And if the customer prefers talking to a live agent, AI can assist the agent by sharing similar information to help personalize the experience. In my book, I’ll Be Back, I write about the Cheers Factor, a callback to the 1980s sitcom with the theme song’s memorable line, “You want to go where everybody knows your name.” This is what customers want: to be noticed, recognized, and appreciated. An AI-assisted personalized experience takes knowing your customer to another level. Consider this: Memory is a loyalty multiplier!

The future of customer experience isn’t AI versus humans. It’s AI working for humans. When AI handles the repetitive work, remembers customers, and knows when to step aside for a live agent, companies deliver faster service, more personalized experiences, and greater confidence that gets customers to say, “I’ll be back!

Image Credits: Gemini

Subscribe to Human-Centered Change & Innovation WeeklySign up here to join 17,000+ leaders getting Human-Centered Change & Innovation Weekly delivered to their inbox every week.

Is the choice really ours?

Is the choice really ours?

GUEST POST from Geoffrey Moore

This is post #4 in my trek to address Twenty Philosophical Questions That Are Hard to Wrap Your Head Around. Here is the blurb that introduces it:

The eternal debate between determinism and free will continues to captivate philosophers as they explore the boundaries and implications of the agency. The philosophical dilemma surrounding will and determinism revolves around the existence of will in the context of determinism. Therefore, it is essential to establish an understanding of what we mean by ” will” and “determinism.” There are generally two schools of thought, among philosophers when it comes to the concept of will. Compatibilists argue that free will simply means being able to do what we want without any form of coercion – in other words, actions. Incompatibilists believe our desires can be manipulated and argue that true free will requires having a choice. They believe that a choice is genuine only if we could have chosen differently if allowed to make the choice again.

I’m not sure who does these blurbs — maybe an early version of ChatGPT? — but this one really does miss the mark. The primary issue to address here is not will; it is determinism. And this is an area where Western thought has wandered far astray, beginning with the thought experiment called Laplace’s demon, summarized here by our good friends at Wikipedia:

According to determinism, if someone (the demon) knows the precise location and momentum of every atom in the universe, their past and future values for any given time are entailed; they can be calculated from the laws of classical mechanics.

This wasn’t even true in Newton’s universe, and it is certainly not true in Einstein’s. The fact is, as chaos theory, emergence, and self-organization have made eminently clear, strict determinism is simply bunk. Yes, you can make a reasonable case for deterministic cause and effect within the boundaries of any particular level of reality, any specific stair in the Infinite Staircase, but the whole point of emergence is that causal relationships are disrupted at the transition between any two stairs. Some effects cross over, but others do not, and some phenomena emerge that are not predictable from the underlying factors. In short, we live in a world that is not just in flux; it is continually presenting layer after layer of novelty, both to us and within us.

In this context, the interesting questions are: to what degree are events, and in particular human actions, genuinely determined? And, to the degree that they are, what issues does this raise for self-determination, free will, and social accountability? This is a subtle problem, and we need to approach it with care.

Start with genetics. The notion that your genes irreversibly determine your biological fate is way too simplistic. Yes, they have impact, but in highly complicated and unpredictable ways, and so do a host of other factors, including diet, exercise, physical environment, emotional state, as well as epigenetics, which entail changes to the genome that emerge during life and are not inherited. In short, we are in the domain of probabilities, not deterministic causes and effects, so yes, we still want to be informed about our genome, but with a relatively modest set of stark exceptions, we should not feel imprisoned by it. Rather, we should treat it instead of the hand we have been dealt and position ourselves to play it to best advantage.

The same principle holds true for the social situation we inherit along with our genes—our parents, our siblings, our ethnicity, the era and place into which we are born, our social and economic class, our education, our job history, not to mention all the chances that simply come our way both good and bad—these are all material factors that are not under our control. All these factors do shape our choices, and they affect the probability of our success in life’s endeavors—but it is a huge overstatement to say they determine them. There are just too many examples of people who have overcome extraordinary adversity, not to mention people who have totally screwed up what was a highly privileged starting position.

Why is this so important? It comes down to responsibility and accountability, most obviously to others, but most importantly to ourselves. Let’s start with others first. As I discuss at some length in The Infinite Staircase, there is a tug of war in contemporary liberal democracies between legal justice and social justice, especially when it comes to dealing with crimes committed by people who are deeply disadvantaged. The progressive position in this tug of war rotates heavily to empathy for these folks and away from accountability. As the city management challenges that have emerged in recent years in Seattle, Portland, and San Francisco have made clear, however, that just doesn’t work. Social cohesion is anchored in rule of law. To be sure, there is every reason to reach out to disadvantaged people with a helping hand, but it does not work to hand out a free pass.

The narrative that supports free passes is called the oppressed versus the oppressors. It argues that you cannot punish the oppressed for actions they are committing in resistance to their oppressors. On the other side, the narrative that supports strict law enforcement is called civilization versus the barbarians. It argues that law is all that keeps us from endless violence against each other. Both narratives at their heart are self-serving, and neither speaks directly to the single most important issue, namely how accountable should we, each of us as individuals, be to maintaining our own welfare?

The answer at one level is pretty simple. We have to take responsibility and accountability for ourselves because no one else can. We are the only ones on site all the time. It is our life, not someone else’s. It is a hand that we have been dealt, and no one else can play it. This is true regardless of how advantaged or disadvantaged we are. In a very real sense, this is not fair, but it is also not negotiable.

So, in that context, now let us ask the question, are we truly free to make choices? It’s actually kind of a weird question. We have to make choices. We have no choice but to make choices. Life demands that we have a strategy for living, and for better or worse, that is what we act out in our daily lives. So, let’s take that as a given and address instead what I take to be the real question on the table: To what degree are we free to determine the choices we make?

I don’t want to be flippant here (well, maybe I do, but I shouldn’t), but to quote Bill Murray from the movie Meatballs, “It just doesn’t matter!” As someone asked the card player why in the world would he play in a crooked game, the answer was because it’s the only game in town. As long as we choose to live, we cannot fold the hand we are dealt, but we do get to discard and draw new cards from time to time, sometimes to good effect, sometimes not. Sometimes this feels free, sometimes it feels forced, and always, it is being influenced by our history and circumstances. But so what?

I give the Existentialists a lot of credit for how they took on this challenge. Their strategy for living is to commit to a set of values, often embodied in a life narrative, and then remain as true to them as possible. Their goal is to be authentic. In itself, that doesn’t mean too much, as you can be authentically pretty horrible, but when it is combined with values that are true to our mammalian roots, it creates a powerful stabilizing force to help meet the buffeting winds of fate.

Being authentic means taking responsibility for our choices, regardless of how free or not they are. That in itself is liberating. It can free us from narratives of victimhood that can be alluring in the moment but devastatingly destructive in the long run. Most importantly, it can support a strategy for living that is coherent and reliable, something others can trust and we can build on.

That’s what I think. What do you think?

— Image credit: Pexels

Subscribe to Human-Centered Change & Innovation WeeklySign up here to join 17,000+ leaders getting Human-Centered Change & Innovation Weekly delivered to their inbox every week.

Microsoft Ignite 2026: Managed Intelligence and the AI Adult

Microsoft Ignite 2026: Managed Intelligence and the AI Adult

by Braden Kelley and Chateau G Pato


Every few years, a technology industry that has been sprinting stops, looks around and discovers that someone needs to do the paperwork. Cloud had its moment. Mobile had its moment. Now AI is having it.

At Ignite this week, Microsoft made a play to be the company that does the paperwork. The pitch, as I read it, is a managed intelligence future: companies use the best frontier models and Microsoft’s own focused models, keep control of security and governance, watch their spend, protect their intellectual property, and tune models and agents to their own work. The tone was less “look what AI can do” and more “let’s be responsible about what we’ve started.” Call it the adult-in-the-room positioning. Whether Microsoft has earned that chair is the interesting question.

I read through Microsoft’s announcements, the analyst commentary and the competition. Here’s what’s new, where the next innovations are, who else is already building this, and where Microsoft looks strong or exposed.

The short answer

Microsoft has assembled the most complete packaged version of managed intelligence I’ve seen from one vendor: its own lower-cost models, a router that decides which model to use, tuning tools, an identity-based control plane for agents, security agents, and a cost dashboard. But much of it is preview, most of the performance claims are Microsoft’s own, and the tuning approach creates switching costs. Others, notably ServiceNow, Google and AWS, are building pieces of the same idea. Nobody I found matches the full stack from chips to cost dashboard.

What Microsoft announced

Smaller, focused models instead of one model for everything

This is the idea I like best, and it matches what I wrote in the hybrid intelligence piece: most work doesn’t need the biggest brain in the building.

  • MAI Code 2 targets everyday enterprise coding. Microsoft says it reaches 98% of GPT-6.1 Sol’s quality at about 40% of the price, and 96% of Claude Opus 5.5’s quality at about 5% of the price, averaged across four coding benchmarks.
  • MAI Cyber 2 Flash is a security model that Microsoft says scores 92.3% on a cybersecurity benchmark against 91.5% for a leading frontier model, at under 2% of the cost per task.
  • Microsoft-Decision-1.1 is a routing model that sends each request to a larger model only when it’s needed. It’s generally available.
  • Frontier models aren’t going away. GPT 6.2 and Meta’s latest models are coming to Foundry, and in Copilot, users can pick OpenAI or Anthropic models, subject to organizational policy.

Microsoft’s own tagline for the lineup: “Frontier performance. A fraction of today’s prices.” Hold that thought, because every one of those benchmark figures is vendor-reported.

A control plane for agents

Agent 365 is the center of the governance story. Microsoft says more than 50 million agents are already in it. Each agent gets its own Microsoft Entra identity, so IT can see it, limit it and block it. Autopilot, a persistent agent that works across Teams and Outlook, comes with its own agent ID and a named human manager, and its actions are logged under its own name for audit.

The most interesting piece is the money. Agent 365 Cost Management tracks consumption by user, service, agent and model, with limits, policies and alerts. Business Analytics measures value “in work completed, not simply tokens.” Satya Nadella called it an “ERP system for AI usage,” and I think that’s the right metaphor. AI spend is about to need the same discipline as any other enterprise resource. Both pieces are in preview or the Frontier program, with value insights expanding in December.

Security that’s agentic too

Microsoft Perception, now generally available, uses red (attack), blue (detection) and green (remediation) agent teams. Defender runtime protection for local AI agents is also generally available, and Constellation Research notes that Microsoft’s security coverage extends to AWS and Google Cloud. Azure confidential computing previews, including a Confidential Clean Room for multiparty analytics, speak directly to the IP-protection concern. Microsoft also announced an AI Transparency Initiative with 40 participating organizations and work alongside NIST’s agent standards effort.

Tuning, so the model learns your business

  • Training in Foundry (preview) adapts supported models with an organization’s own data. Programmed Training (preview) lets engineers write their own training loops for supervised and reinforcement learning on open-weight models. The agent optimizer, which proposes and tests agent changes against evaluations, is generally available.
  • Frontier Tuning, announced at Build in June and in private preview, uses reinforcement learning inside the customer’s compliance boundary to learn from process, not just examples. Microsoft’s reported results include an HR workflow going from 13% to 87% task completion and a tuned model at roughly one-tenth the cost for a McKinsey use case. Those are vendor-reported figures, with no methodology.
  • MAI Licensed Weights would give trusted customers access to Microsoft’s model weights. No date has been given.

The context layer and the hardware

Microsoft IQ packages organizational context, drawing on Work IQ, Fabric IQ, Foundry IQ and Web IQ, with about 350 third-party connectors. Constellation’s summary of the strategy is that models are a commodity and the value is in the context. On the hardware side, Microsoft claims its Maia 300 chip delivers 40% more tokens per watt than leading GPU-based systems, and Cobalt 200 virtual machines are generally available. For customers who want the work closer to home, Foundry Local, Windows Server 2028 (preview) and Azure Local options address sovereignty and residency.

Why this matters

Three shifts here are worth naming, because they’re bigger than any single product:

  1. Routing is becoming a management function. Deciding which model handles which task, based on quality, speed and cost, is the AI equivalent of deciding who on your team should take a call. It’s an operating decision, not a technical one.
  2. Cost is finally a first-class metric. Measuring value in work completed rather than tokens consumed is the correct instinct. Counting tokens is like measuring a salesperson by miles driven.
  3. Control is moving from the model to the agent. Giving each agent an identity, a manager and an audit trail treats agents like a workforce, which is how people will need to think about them.

Where the next innovation is

This is the part I’d bet on. The announcements open gaps, and gaps are where innovators live.

  • Portability of tuned intelligence. If a model has been tuned on how your company actually works, that asset is valuable. Whoever makes it portable, auditable and exit-friendly wins the trust of every skeptical CFO.
  • Independent benchmarks of cost and quality. Today the buyer is comparing one vendor’s slides to another’s. A neutral, auditable cost-per-outcome standard would be worth real money.
  • A cross-vendor control plane that’s actually neutral. Companies will run models from several providers. The control layer that works across all of them, without favoring its owner, is still up for grabs.
  • Outcome accounting. Measuring “work completed” is hard. The first tools that tie agent activity to business results, not activity logs, will change how AI budgets get approved.
  • The human side of an agent workforce. Nobody is yet systematically measuring how it feels for employees and customers when agents join the process. That’s an experience problem, and it’s why I believe experience audits will extend to the human-and-agent journey.
  • Scenario planning for model prices. If specialized models cost a fraction of frontier ones, your 2027 budget assumptions may be wrong in either direction. That’s a signals-and-futures exercise, not a spreadsheet exercise.

Who else is doing this?

Yes, others are already building parts of it. The market is crowded, and its claims are often ahead of its evidence. Here’s what I found, with sources noted below. These are vendor descriptions unless I say otherwise.

Vendor What it offers Where it looks strong Where it looks exposed
ServiceNow (AI Control Tower) Discover, observe, govern, secure and measure AI across systems, including AWS, Google Cloud, Azure, SAP, Oracle and Workday; cost and ROI dashboards; a “kill switch” Governance that turns into real work: routing findings to owners, tracking remediation, keeping an audit trail One analyst survey found 14 of 15 customers said runtime control was external, limited or unproven at scale; 12 of 15 raised cost concerns
Google (Gemini Enterprise Agent Platform) Agent identity, registry, gateway, evaluation and observability; Model Armor; 200+ models in Model Garden; FinOps Explainability agent and project spend caps Governance built into the platform, open at the edge and consolidated at the center, per Bain Centered on Google Cloud; maturity of each component unclear; tuning wasn’t addressed in the coverage I read
AWS (Bedrock AgentCore) Cedar-based policies that check tool calls, temporal policies for approvals and budgets, evaluations (GA March 2026); model customization including reinforcement fine-tuning for Nova The closest to Microsoft’s full-stack idea: its own models, customization and agent policy in one provider AWS is the control surface; policy covers traffic through its gateway; compliance mapping wasn’t documented
IBM (watsonx Orchestrate, Sovereign Core) An agent control plane for multi-cloud and on-premises, with sovereignty controls Strong on hybrid and regulated environments; Sovereign Core is generally available The next-generation Orchestrate was in private preview as of May; cost controls and observability weren’t documented on its pages

A few caveats. Much of the head-to-head comparison comes from a guide published by a vendor that sells a competing product, which it discloses, so I used it for feature facts and not for rankings. I didn’t dig into Salesforce, Databricks or the smaller governance specialists, so this table isn’t a market map. And no one else I found combines their own cheaper models, routing, tuning, governance and silicon the way Microsoft is pitching. AWS comes closest.

Microsoft’s strengths

  • It already owns the identity layer. Governing agents through Entra, Purview and Defender builds on tools many enterprises already run. That’s a real distribution advantage.
  • Full-stack cost control. Few competitors can say “our model, our router, our chip and our cost dashboard” in one sentence.
  • Multi-model by design. Letting customers pick OpenAI or Anthropic models, subject to policy, reduces the “all eggs, one basket” objection.
  • A credible narrative. Measuring value in work completed, and framing AI spend like an ERP discipline, speaks the language CFOs and boards already use.

Microsoft’s weaknesses

  • Lock-in through tuning. One critique of Microsoft’s approach calls the “no vendor lock-in” claim “technically correct and commercially misleading at the same time.” Models tuned on your workflow traces become sticky, and leaving means retraining elsewhere. Microsoft says the knowledge stays yours, but the coverage doesn’t address portability of the tuning environment or skills.
  • Unproven governance across mixed estates. Entra is mature for people. Governing agents across Azure, Google Cloud and on-premises systems is newer, and ServiceNow’s tool was designed from the start for agent governance.
  • Marketing numbers. The cost-versus-quality comparisons are Microsoft’s, and its Copilot announcement cites a GigaOm report claiming Copilot is 84% cheaper than Claude for an everyday user, which I couldn’t verify independently. Parity today doesn’t mean parity in six months, as the releases keep coming.
  • Licensing complexity. Insights for Microsoft services require at least one Microsoft 365 Copilot license in the tenant, and insights for non-Microsoft agents require an Agent 365 license. A third-party guide lists Agent 365 at $15 per user per month, best with E5. That’s a platform commitment, not a plug-in.
  • Lots of preview. Cost Management, Frontier Tuning, Programmed Training and MAI Licensed Weights are preview, private preview or undated.
  • Reliability. A global Microsoft 365 outage knocked out Outlook, Teams and OneDrive for nearly 24 hours last week, and the UK’s NHS lost email and Teams access for much of the day. Microsoft apologized, attributed it to a network change, and said it wasn’t related to AI or agents. Fair enough, but it’s hard to be the adult in the room while the lights are out.
  • The IP promise. Microsoft has promised “enterprise grade, clean and commercially licensed data.” A June report from The Decoder said Microsoft’s own technical paper shows a significant portion of the MAI training corpus comes from Common Crawl, an openly available web scrape. I didn’t see a Microsoft response, and the report doesn’t give a proportion. Buyers who care about IP protection should ask for the training-data details in writing.

What I’d ask before signing anything

  1. Can I take my tuned model, skills and evaluations with me, and what does it cost to leave?
  2. Who verified the benchmarks, on whose workloads? Run your own tasks before you believe anyone’s price-performance chart.
  3. What does it cover outside this vendor’s walls? Ask to see agents from other platforms governed, blocked and logged in a live demonstration.
  4. What’s generally available today, and what’s a roadmap?
  5. How is value measured? If the answer is “tokens,” keep asking.
  6. What happens to my employees’ and customers’ experience? Measure it before and after.

My verdict

Innovation or Not? Innovation, mostly in the packaging, with proof still to come.

None of the individual ideas is brand new: model routing, agent identity, FinOps and fine-tuning all exist elsewhere. What’s new is a large vendor assembling them into a single story that a CIO, a CFO and a general counsel can all understand, and putting its weight behind the idea that AI needs management, not just momentum. That’s a meaningful shift, and it will force competitors to answer.

But being the adult in the room is something you demonstrate, not something you announce. The test will be portability, independent verification and keeping the lights on. If Microsoft passes those, managed intelligence becomes the default way large organizations buy AI. If it doesn’t, the opening goes to whoever builds the neutral control layer.

Either way, the question for leaders has changed. It’s no longer “should we use AI?” It’s “who is managing it, and how would we know?”

Sources

Image Credits: Microsoft AI

Content Authenticity Statement: The topic area, key elements to focus on, etc. were decisions made by Braden Kelley, with a little help from Claude to clean up the article.

Subscribe to Human-Centered Change & Innovation WeeklySign up here to get Human-Centered Change & Innovation Weekly delivered to your inbox every week.

Nvidia’s Earnings Beat the Drama — What Leaders Should Watch Next

Nvidia’s Earnings Beat the Drama — What Leaders Should Watch Next

by Braden Kelley and Art Inteligencia


What Did Nvidia’s Earnings Show After Jensen Huang Stepped Down? (Short Answer)

On November 17, 2026, Nvidia reported third-quarter fiscal 2027 revenue of $115.8 billion — up 103% year over year — with Data Center at $107.6 billion and non-GAAP diluted EPS of $2.68, beating Wall Street expectations. Guidance for the current quarter centers on $130 billion (±3%), with zero Data Center compute revenue from China assumed. This was the first report under interim CEO Colette Kress, days after co-founder Jensen Huang resigned; the board said it will not reinstate him and is searching for a permanent successor. The numbers say demand is still climbing. The open questions are whether brighter scrutiny on diversion and export paths will clip growth — and whether open-source AI and physical AI expand the ceiling faster than geopolitics can shrink it.

Leadership drama changes the face of a company. Demand for compute changes the future of every company that needs intelligence — digital or physical.

What Are the Nvidia Earnings Highlights Leaders Should Actually Remember?

Strip away the noise and the quarter is still extraordinary:

  • Revenue: $115.8 billion, up 20% quarter over quarter and 103% from a year ago.
  • Data Center: $107.6 billion, up 110% year over year — still the engine.
  • Edge Computing: $8.2 billion, up 42% year over year — PCs, local AI, automotive, and robotics-adjacent workloads showing through.
  • Profitability: GAAP diluted EPS $2.70; non-GAAP $2.68. Gross margins roughly mid-70s this quarter, with Q4 guided lower into the low-70s as the mix and costs evolve.
  • Vera Rubin: Nvidia began shipping Vera Rubin NVL72 — called the fastest product ramp in company history — into major clouds and AI infrastructure partners. Call coverage put first-quarter production shipments for the platform near $20 billion.
  • Capital return: The board had authorized a $150 billion increase to the share-repurchase program; Nvidia returned $28.1 billion to shareholders in the quarter via buybacks and dividends.
  • Forward look: Q4 revenue guided to ~$130 billion (±3%). Kress pointed to roughly 75% year-over-year growth into fiscal 2028 — higher than the prior ~70% framing — and argued demand is still accelerating as supply catches up.

For a company that just lost its legendary founder-CEO in a public storm, that is not a soft landing for the spreadsheet. It is a hard reminder that the AI factory buildout did not pause for the news cycle.

Outstanding results do not erase risk. They buy time — and raise the cost of getting the next chapter wrong.

Will Export Scrutiny and Diversion Risk Cut Nvidia’s Sales?

Here is the fear in plain language: if regulators shine a brighter light on chips sold to countries and entities that might divert them toward restricted Chinese end users, will the top line crack?

Nvidia is already running the P&L as if China Data Center compute is not a planning assumption. That is not new theater — it has been the guidance posture for several quarters — but the context around Huang’s resignation has intensified national-security attention, including tighter rules aimed at rental and third-country pathways that could circumvent export controls. Diversion risk is real. Licensing friction is real. Sovereign customers will keep asking harder questions about where silicon ends up.

That can slow some deals, raise compliance cost, and force uglier channel choices. It does not automatically equal a collapse in demand. A company guiding toward $130 billion in a single quarter while assuming zero China Data Center compute is telling you something important: the rest of the world is still starving for accelerators. Export policy can reshape who buys, how they buy, and at what delay. It has not yet proven it can delete the underlying shortage of useful compute.

The leadership test for Kress — and whoever becomes permanent CEO — is not clever denial. It is building a compliance and partner architecture that is boring enough for Washington and still fast enough for customers who refuse to wait.

Geopolitics can redirect demand. It rarely invents or destroys the need for intelligence at scale.

Is Compute Demand Still Underestimated — Especially for Physical AI?

Most public debate still frames AI demand as chatbots and hyperscaler training runs. That story is incomplete.

On the call and in follow-on coverage, Nvidia pointed to open models and agentic workloads as major token drivers, enterprise and financial-services on-prem growth, sovereign AI projects that more than tripled, and physical AI revenue that nearly tripled. Edge Computing’s climb, Isaac robotics advances, Cosmos world models, automotive partnerships, and humanoid reference designs are not side hobbies. They are early receipts for a world where intelligence leaves the data center and enters factories, vehicles, warehouses, and homes.

Digital agents already multiply inference. Physical AI multiplies sensors, simulation, on-device inference, fleet learning, and safety validation. If robotics and autonomous systems scale even halfway toward the hype, today’s “massive” data-center demand is a floor, not a ceiling — and many forecasts still underweight that second curve.

None of this means every GPU sold this year becomes a humanoid tomorrow. It means leaders who plan only for LLM training peaks will misread the next decade of infrastructure, energy, and talent. Soft landing for organizations: treat physical AI as a portfolio of experiments with real operators. Hard landing: wait until a competitor’s robots, vehicles, or plants already run on a stack you cannot catch.

Chat needs compute. Bodies that sense and act need more — and they need it closer to the work.

What Does a Post-Huang Nvidia Mean for Open-Source AI and Innovation?

Huang was the evangelist-in-chief — the person who made AI infrastructure feel inevitable. Kress is the operator who has translated that story into numbers for thirteen years. Different strengths. Different risk: vision without governance, or governance without ambition.

The open-source and “open platform” signals in this report matter precisely because the permanent CEO seat is still open:

  • Vera Rubin open architecture advances, with major cloud and infrastructure players joining to help the industry scale faster.
  • NVIDIA Open Agent Safety Platform and an OpenShell reference architecture that mixes open-source software with silicon-enforced monitoring of agent behavior.
  • Continued push on Nemotron open models and developer ecosystems — including claims of strong open-model usage on routing platforms.
  • An agreement to acquire Hugging Face — a landmark bet on making models, datasets, and community infrastructure more open and accessible at global scale.
  • Open robotics and research stacks around Isaac / GR00T-style platforms — openness as a way to seed an ecosystem Nvidia still supplies with silicon.

That is not altruism. It is platform strategy. Open models and open tooling expand the number of builders who need accelerators, networking, and AI factories. Closed-only stacks invite alternatives. The irony of this moment: the founder who personified Nvidia’s rise is out, while the company doubles down on openness as a growth engine. The next CEO will decide whether Hugging Face, open agents, and open robotics stay at the center — or get managed like a branding sideline when margins tighten.

For innovators outside Nvidia, the lesson is older than this earnings release. Open innovation works when boundaries, IP rules, and adoption paths are clear — and fails when “open” is a portal with no owner. Nvidia’s bet is that owning the rails of open intelligence still beats trying to own every model. Whether that remains true under permanent leadership is now a foresight question, not a cult-of-personality question.

Open source without a path to adopted value is theater. Open platforms with silicon, safety, and distribution can rewrite who gets to innovate.

What Should Leaders Watch Between Now and the Permanent CEO?

  1. Compliance vs. growth tradeoffs — Does tighter diversion enforcement change product mix, partner list, or delivery timelines more than guidance admits?
  2. Vera Rubin adoption quality — Shipments are one signal; sustained utilization and customer concentration are another.
  3. Physical AI conversion — Which robotics, AV, and industrial deployments leave the demo and enter fleets?
  4. Hugging Face integration — Does the acquisition accelerate open innovation — or trigger trust and competition backlash that fragments the ecosystem?
  5. CEO choice — Operator, product visionary, or geopolitical diplomat? The permanent hire will reveal which future Nvidia thinks it is in.

Huang has said he would return if asked. The board has said it does not intend to reinstate him. Markets will gossip. Strategists should map scenarios: continued Kress-led execution, an internal product successor, an external geopolitical CEO, or a hybrid board that separates face, factory, and Washington. Preferable futures are designed — not inherited from a petition or a headline.

The earnings answered whether demand survived the week. They did not answer who will steer the decade.

Ready to Take the Next Step?

Nvidia’s quarter is a reminder that the future of AI will not be decided by one earnings print or one CEO transition. Export rules, open models, agent workloads, and physical AI are signals you can explore, prepare for, and shape — instead of treating them as weather. That is why I created FutureHacking™, a strategic foresight methodology that helps individuals and teams become their own futurist.

FutureHacking™ gives you a simple four-step approach, supported by 20+ visual, collaborative tools:

  1. Pick the signals that matter — compute scarcity, export controls, open-source platforms, and physical AI, not only the celebrity CEO story.
  2. Map how those signals evolve — so you can tell a passing headline from a real change.
  3. Explore possible, probable, and preferable futures — instead of betting everything on a single prediction about Nvidia or China.
  4. Make your preferable future a reality — with a roadmap, guideposts, and actions you can start now.

Whether you are planning your own career in an AI-shaped economy or helping a leadership team decide where to place infrastructure and innovation bets, FutureHacking™ turns market drama into a decision system.

Learn more about the FutureHacking™ methodology →

FAQ: Nvidia Earnings, Leadership, and AI Demand

What were Nvidia’s Q3 fiscal 2027 earnings?

Nvidia reported revenue of $115.8 billion for the third quarter of fiscal 2027, up 103% year over year, with Data Center revenue of $107.6 billion. Non-GAAP diluted earnings per share were $2.68. The company guided fourth-quarter revenue to about $130 billion, plus or minus 3%.

Who is Nvidia’s CEO after Jensen Huang?

Colette Kress, Nvidia’s longtime CFO, is serving as interim CEO after Jensen Huang resigned in November 2026. The board said it does not intend to reinstate Huang and is conducting a search for a permanent CEO.

Is Nvidia still selling AI chips related to China?

Nvidia’s outlook assumes no Data Center compute revenue from China. Export controls, diversion concerns, and tighter rules on third-country and rental pathways remain active risks that can reshape who can buy advanced chips and how.

Why might Nvidia’s compute demand keep rising?

Beyond large language model training, demand is broadening into agentic workloads, enterprise and sovereign AI, and physical AI — robotics, autonomous vehicles, and industrial systems that need simulation, onboard inference, and fleet learning. Those curves can keep accelerator demand elevated even when one geography is constrained.

What does Nvidia’s Hugging Face deal mean for open-source AI?

Nvidia announced an agreement to acquire Hugging Face to scale an open platform for models and community infrastructure. Combined with open architecture, open models, and open agent-safety tooling, it signals that openness is part of Nvidia’s growth strategy — though execution under a permanent CEO will determine whether that bet deepens or narrows.

Image Credits: Cursor

Content Authenticity Statement: The topic area, key elements to focus on, etc. were decisions made by Braden Kelley, with a little help from Cursor to clean up the article.

Subscribe to Human-Centered Change & Innovation WeeklySign up here to get Human-Centered Change & Innovation Weekly delivered to your inbox every week.

Data Centers in Space – Innovation or Not?

Data Centers in Space - Innovation or Not?

by Braden Kelley and Art Inteligencia


Somebody has to say it out loud: the plan to put AI data centers in orbit sounds like something I’d have pitched as a joke in a brainstorm, right before somebody reminded me that radiators are heavy.

And yet Google has TPUs in orbit. SpaceX has asked the FCC for permission to launch up to a million computing satellites. Blue Origin wants 51,600. A startup called Cowboy Space wants to turn the upper stage of its own rocket into a megawatt data center. Starcloud has already flown an Nvidia H100.

Two questions are worth asking, and they’re easy to blur together. Does it make economic sense? And is this really happening because Americans are turning against data centers on the ground? I’ll take the second one seriously, because the politics are real. But I’ll start with the money.

The short answer

Not today, and probably not this decade on cost alone. Orbital compute carries a large price premium over ground-based compute, and the best independent analysis I found expects that premium to shrink but not vanish. The case for space is less “it’s cheaper” than “it’s available.” That’s a different argument, and a more interesting one.

And on the politics: every major orbital data center filing I found was made before the November 3 midterms. The idea didn’t come from the election. The election just made the idea easier to sell.

Who’s building what

Here are the five efforts I’d watch, side by side. Where I’ve given a figure, it comes from the company’s filing, announcement or press coverage, not from a flight record.

Approach Scale they’ve asked for What’s actually flown Next milestone
Google (Project Suncatcher) Small satellites with TPUs, flying in tight formation and linked by lasers, in near-constant sunlight An illustrative cluster of 81 satellites Four TPUs on a Planet-built prototype, launched October 1, 2026 In-orbit tests now underway
SpaceX Massive constellation of AI satellites riding on Starlink’s laser network and Starship launches Up to 1,000,000 satellites (FCC filing) Nothing yet that I could verify Deutsche Bank expects small tests in 2027 or 2028
Cowboy Space Rocket upper stage is the data center: a 1 MW node with five server racks Up to 20,000 satellites (FCC filing, pending) Reason-1, a power-beaming demo launched October 1, 2026 First Big Iron rocket launch targeted for 2028
Blue Origin (Project Sunrise) Satellite constellation tied to its TeraWave network Up to 51,600 satellites (FCC filing) Nothing yet that I could verify No schedule given; seeking waivers of FCC deadlines
Starcloud Modular satellites built around Nvidia chips, launched mostly on SpaceX 88,000 satellites and 20 GW (FCC application) Starcloud-1 with an Nvidia H100, launched November 2025 Two 8 kW satellites, first one January 2027

Google: the careful engineers

Google’s approach reads like a research paper because it is one. The satellites would fly in a dawn–dusk sun-synchronous orbit, where the panels see near-constant sunlight. Google says panels in the right orbit can be up to eight times more productive than on Earth. Its team tested a Trillium TPU in a proton beam and found the memory showed irregularities only after about three times the dose expected over five years of shielded orbit. Google called the chip “surprisingly radiation-hard for space applications.”

Google also did the most useful thing any of these players did: it published the economics. Its analysis suggests launch prices could fall below $200 per kilogram by the mid-2030s, which is where it says space-based data centers become roughly cost-comparable to terrestrial energy costs. Notice how specific that is. It’s a condition, not a promise. And Google itself says “significant engineering challenges remain,” including thermal management and reliability.

The prototype launched on October 1, 2026 on a SpaceX rideshare, and Google reported it was operating as expected. A trade-press roundup from early November says the prototype ran Gemma 4 on an orbital TPU, but only for about 20 minutes of one TPU at a time because of power limits. That’s a proof of concept, not a data center.

SpaceX: the biggest ambition and the loosest details

SpaceX’s FCC filing asks for approval of up to one million satellites at altitudes of 500 to 2,000 kilometers. The filing’s implied math is roughly a million tonnes of hardware per year, adding about 100 gigawatts of AI capacity annually. It gives no satellite mass, no hardware design and no unit cost. One analyst called it “quite rushed and to be very early stage.”

Then there’s the gap between the pitch and the paperwork. Elon Musk has said space would be the cheapest place to run AI within 30 to 36 months, adding “It’s not even close.” But SpaceX’s own IPO filing is far more modest. It says the orbital initiatives depend on “unproven technologies, or technologies that do not exist,” that they “may not achieve commercial viability,” and that the timeline for 100 GW “may be difficult or impossible to determine.”

The launch arithmetic explains the caution. 100 GW is on the order of 10,000 Starship launches. the economics lean on fully reusable Starship boosters, which an early-2026 analysis said SpaceX had not yet demonstrated. And the filing says SpaceX does “not anticipate servicing or repairing processors in space,” so every failed chip is a lost chip.

Credit where due: SpaceX has a record of doing the improbable. The Colossus cluster, from xAI, scaled to 100,000 GPUs in 122 days against a 24-month estimate. That’s a reason not to dismiss it. It isn’t evidence that orbit works.

Cowboy Space: the most original architecture

Cowboy is the one I find most interesting, because it attacks a different bottleneck. Most orbital plans launch a satellite on someone else’s rocket and discard the upper stage. Cowboy’s idea is to keep the upper stage, make it the data center, and skip the redundant structure so more of the launched mass goes to compute, power and cooling.

The company was formerly Aetherflux, a space-based solar power startup, and was founded in 2024 by Baiju Bhatt, a Robinhood co-founder. It raised a $275 million Series B led by Index Ventures at a reported $2 billion valuation. Its Stampede constellation is filed at up to 20,000 satellites, each a 1 MW node with five server racks, using Nvidia’s Space-1 Vera Rubin modules. The rocket, Big Iron, is an 86-meter stainless-steel vehicle with nine reusable methalox engines. The first full vehicle is targeted for the end of 2027 and a maiden launch for 2028.

The pitch is blunt: “Earth’s energy grid can’t run at the pace of AI. We can.” The company also says average grid connection lead times for new data centers in major U.S. markets “can run five to seven years, or more.” I haven’t verified that figure independently, so treat it as the company’s claim.

The trade-off is concentration risk. One analysis of the sector put it plainly: smaller nodes lose little when one fails, while Cowboy’s larger nodes lose far more, and the company has to build a new rocket and win FAA authorization first. Its own site is candid that its capabilities and timelines are “design objectives and goals, not demonstrated or guaranteed outcomes.” And the FCC sent a deficiency letter in August asking about satellite dimensions and debris mitigation.

The rest of the field

Blue Origin has asked the FCC for up to 51,600 satellites under Project Sunrise, and Jeff Bezos has said that “We will be able to beat the cost of terrestrial data centers in space in the next couple of decades.” Note the timeframe: couple of decades. That’s a more candid number than the ones coming out of SpaceX. Starcloud is the most flown, with an H100 in orbit since November 2025, and has raised about $450 million at a $2.3 billion valuation. Its CEO, Philip Johnston, concedes that if he can’t book SpaceX launches in 2029, “that will be challenging for us.” Which tells you who the whole industry is renting from. Orbital Compute, a Los Angeles startup with a $5 million pre-seed, is proposing up to 100,000 modular 100 kW satellites.

Is it cost effective?

The best independent number I found comes from Boston Consulting Group. Its conclusion: orbital facilities carry a 2.5x to 3x cost premium today. Over 20 years, BCG puts the cost at roughly $660 million to $750 million per megawatt in orbit versus $230 million to $300 million on the ground. Its closing line is one I’d put on a slide: “Space-based data centers won’t win on cost—but they don’t have to.”

BCG’s improvement scenarios are worth reading closely:

  • A realistic 5-to-10-year path (launch costs down about two-thirds, satellites half as heavy) leaves a premium of roughly 1.5x to 1.8x, depending on how many satellites fail.
  • An aggressive path (full Starship-level economics) could get the premium down to 1.1x to 1.2x, but only if failure rates stay very low.
  • Parity this decade is judged unlikely.

At roughly $1,500 per kilogram, BCG says putting 1 GW of compute in orbit would cost around $30 billion in launch alone, which it calls economically impractical. A separate open calculator built by Andrew McCalip, which Tom’s Hardware covered, found a baseline of about $51 billion for a space installation against $16 billion on the ground. Its author’s verdict on the whole idea was “FOMO and aesthetic futurism.” That model covers only launch logistics, and he says so.

Three things keep the economics stubborn:

  1. Heat. In a vacuum you can only shed heat by radiation. SpaceX’s own filing says its compute satellites would need “substantially larger radiators” than Starlink’s. The International Space Station sheds tens of kilowatts; an AI cluster has to shed megawatts.
  2. Failures. Chips fail, especially under full load, and nobody is going up to swap a card.
  3. Obsolescence. GPUs age out in a few years. A satellite is a poor place to hold an asset that depreciates that fast.

My own arithmetic on scale makes the point vivid. A single 1 GW campus is a thousand Cowboy-sized 1 MW nodes. Starcloud’s first two flight satellites total 16 kW. That’s the distance between a demo and a business, and it’s why I’d treat any “cheaper than Earth within three years” claim with the same skepticism SpaceX’s own lawyers apparently do.

Is this about the backlash against data centers?

Now to the question you didn’t ask but I know you’re thinking about. Is space only attractive because building data centers on the ground has become politically harder?

The backlash is real, and it’s big. On November 3, voters in 10 states decided 41 local data center measures, and nearly every one limited development. Ohio alone had 21. In Ashville, Ohio, more than 80% of voters rejected a data center already under construction. In Texas, Gina Hinojosa won the governor’s race after calling for a moratorium on new data center approvals. Analysts cited by one trade outlet say up to 60% of planned 2028 U.S. data center projects could face delays, and Goldman Sachs puts 45% of planned capacity at risk. Early this year, at least 75 projects worth about $130 billion were reportedly stalled or blocked.

So yes, Democratic candidates ran on this issue and many won. But if the story is “one party’s policies caused the backlash,” the evidence doesn’t back it up:

  • The opposition crosses party lines. Gallup found about 70% of Americans oppose local AI data center construction, including 75% of Democrats and 63% of Republicans. In Box Elder County, Utah, where Trump took nearly 80% of the vote, a 40,000-acre project drew intense backlash from rural conservative voters over water and tax breaks.
  • Republican governors acted first. Greg Abbott ordered a statewide audit of data centers seeking grid connections, saying “Texans must come first.” Ron DeSantis signed new requirements on hyperscale data centers. Nevada’s Joe Lombardo cut sales-tax benefits and told companies to cover their own power and water costs. Nebraska’s Jim Pillen said “we cannot foot the bill for them.”
  • The concerns are about bills, water and tax breaks. One Federal Reserve Bank of Dallas paper estimated data centers raised wholesale electricity prices 6.9%. Corpus Christi’s reservoir fell to 8% of capacity. Those aren’t partisan facts.

There’s a fair argument on the other side, and I’d make it too. Blanket bans and moratoria are blunt instruments. They can push the same demand to another state or country without fixing the bill problem, and they can slow the very grid and generation upgrades communities need. People who think that way aren’t wrong to worry. The administration has taken that line, with President Trump calling Texas’s opposition a “mistake.”

But here’s my main point. Space doesn’t escape the politics, it changes the venue. Cowboy’s constellation filing is pending at the FCC with a deficiency letter outstanding. Blue Origin is asking for waivers on deployment deadlines. SpaceX wants the standard six-year milestones waived. And the Center for Space Environmentalism has already raised concerns about collision risk, re-entry hazards and interference with astronomy. A million satellites will have their own opposition, and I’d bet it shows up faster than anyone is planning for.

What would have to be true

For orbital data centers to be a real answer, five things need to line up:

  1. Launch costs fall by an order of magnitude, roughly from about $1,500 per kilogram toward the $100 to $200 range.
  2. Heat rejection works at megawatt scale, in hardware that actually flew.
  3. Failure rates stay low enough that losing a satellite isn’t losing a company.
  4. Regulators approve constellations of a size no one has licensed before.
  5. Ground alternatives stay stuck. If efficiency, flexible demand and local small models ease the pressure, the premium stops being worth paying.

That last point is the one I’d emphasize. OpenAI’s Grid Protectors program, which offers up to 1 GW of flexible demand at a planned Georgia campus that utilities could cut in emergencies, is a competing answer to the same constraint. So is running smaller models locally, which I wrote about in the hybrid intelligence piece. Space competes with every one of those, and they’re all available now.

My verdict

Innovation or Not? Innovation, with an asterisk the size of a Starship.

The engineering effort is genuine, the capital is real, and Google’s disclosure of its economics is the kind of honesty this field needs more of. But the evidence says orbital compute is a premium product for buyers who can’t get power on the ground, not a cheaper way to run AI. It won’t relieve the pressure on communities any time soon, because the first real megawatt nodes are a 2028-or-later story and the numbers needed to match a single large campus are enormous.

And the politics? The backlash gave the idea a better story, but it didn’t create it. The filings came first. The more useful lesson for innovators is the one the voters taught: people don’t mind new infrastructure, they mind paying for it, drinking less water because of it, and not being asked. A technology that solves a real constraint and respects the people near it wins. Whether the data center sits in Ohio or in a dawn–dusk orbit, that rule travels with it.

I’d watch two signals: whether the Google prototype’s in-orbit tests show the chips and links holding up, and whether SpaceX’s real launch cadence starts to look like the plan. Those will tell us more than any filing.

Sources

Image Credits: Gemini

Content Authenticity Statement: The topic area, key elements to focus on, etc. were decisions made by Braden Kelley, with a little help from Claude to clean up the article.

Subscribe to Human-Centered Change & Innovation WeeklySign up here to get Human-Centered Change & Innovation Weekly delivered to your inbox every week.

Preparing for Luck is the True Formula for Success

Preparing for Luck is the True Formula For Success

GUEST POST from Greg Satell

One of my favorite podcasts is Derek Thompson’s Plain English and, on a recent episode, he focused on how Stanley water bottles got so popular. It’s a great story that involves a heritage-laden, 166-year-old company, a group of Mormon mothers, astute executives and social media. It’s hard not to listen to the podcast and be inspired to follow in Stanley’s footsteps.

That would be a mistake. As Thompson points out on the podcast and in his book, Hit Makers, runaway successes are unique. They succeed by doing things differently. If everybody followed in the same footsteps, it would be hard for any of them to break away from the pack.

The simple fact is that there is no tried-and-true formula for success. We can compute the odds any way we want, but the truth is that mathematics tells us anything can happen. Take a look at any runaway success and luck played an important part. Yet results aren’t completely random either. We all have luck, both good and bad, the difference is how you prepare for it.

The Strange, Cosmic Quality Of Flukes

I’m of the generation that grew up watching the hit show Happy Days and remember fondly its theme song, “Rock Around the Clock” performed by Bill Haley & His Comets. The song was chosen, I am sure, because as one of the most iconic hits of the 1950s, it epitomized the period that the show was trying to convey.

Yet “Rock Around the Clock” was almost never recorded. Written in 1952, Bill Haley couldn’t get anyone to record it for two years. Its raucous beat just seemed out of step with the clean-cut culture of the early 1950s. Eventually, given a little extra time at the end of a recording session for another song, he was able to fit it in and his record label released it as a B-side.

Then lightning struck. Hollywood star Glenn Ford was looking for a song to reflect the nascent teen revolt he wanted to portray in his upcoming film, Blackboard Jungle. His young son Peter played him “Rock Around the Clock” and it fit perfectly. Blackboard Jungle was a commercial and critical success. Bill Haley’s song shot to the top of the charts and became one of the biggest hits in history.

The incredible success of “Rock Around the Clock” was a fluke, but it wasn’t just dumb luck. Bill Haley was already a successful performer when he recorded it and was encouraged to play other songs. Yet “Rock Around the Clock” was what he wanted to record, over the objections of record producers who wielded power over him. Still, he kept at it.

Without “Rock Around the Clock,” Bill Haley would have been a successful performer. But it was pushing to record a song he believed in that made him a legend.

Einstein, Chance And The Matthew Effect

Although we remember him as an icon today, for a long time, Albert Einstein wasn’t very popular, or even well liked, in the early 20th century. He was German in the wake of World War I, Jewish in an age of heightened anti-Semitism, and so seemingly aloof that he claimed that only a handful of people on earth could understand his strange theories.

That abruptly changed when Einstein first arrived in America on April 3rd, 1921 and a handful of journalists dutifully went to meet him. Yet when they arrived at New York Harbor, they were amazed to find a crowd of thousands waiting for him, screaming with adulation and waving handkerchiefs. Surprised at his popularity, and charmed by his genial personality, the story of Einstein’s arrival made the front page in major newspapers.

It was all a bit of a mistake. The people in the crowd weren’t there to see Einstein, but Chaim Weizmann, the popular Zionist leader that Einstein was traveling with to raise funds for Hebrew University (and who the WASPy science reporters didn’t recognize). Nevertheless, that’s how Einstein gained his iconic status, which would overshadow the other great lights, such as Bohr, Heisenberg and Schrödinger, in the golden age of physics.

From there, the Matthew effect (or what network scientists call preferential attachment) took over. Because Einstein was now so well known, newspapers wanted to report about him and ask him about the other scientific breakthroughs of the day. Just as the rich get richer, the popular get more popular. Einstein became more than a scientist, but a cultural touchstone.

Einstein wasn’t a fluke like Bill Haley. He was already world famous and favored to win the Nobel Prize when he arrived in New York Harbor. Still, there are hundreds of Nobel Laureates that most people have never heard of. Einstein became not only famous, but emblematic of both physics and scientific genius. It’s ironic the trigger came not from doing physics, but by supporting a cause he believed in.

Preparing For Your Trigger

When Vitaliy Shabunin and Dasia Kaleniuk founded the Anti-Corruption Action Center at the height of the Yanukovych regime in 2012, most people thought it was a fool’s errand. Ukraine at the time was hopelessly corrupt. The regime not only stole, it looted with a degree of avarice that was exceptional, even considering the post-communist country’s checkered past.

Yet Vitaliy and Dasia were undeterred. With funds provided from international NGOs, they investigated corruption at state enterprises and possible reforms that could be undertaken. Almost nobody read their reports. They also developed anti-corruption legislation, which no politician would dare to take up. It was grueling, thankless work.

Yet events at the end of 2013 would change the course of history. At the behest of Vladimir Putin, President Yanukovych pulled out of an EU Association trade agreement that he had promised to sign. It led to an outcry which triggered the Euromaidan protests and, in turn, led to the Revolution of Dignity. The regime was overthrown, Russia’s little green men took Crimea and Ukraine, left broke after Yanukovych’s looting, was at war with a superpower.

The powers in the west, such as the US, the EU, the IMF and the World Bank were sympathetic, but were also clear-eyed. They told Ukrainian officials in no uncertain terms that they must install credible anti-corruption reforms or no aid would be forthcoming. The legislation that Vitaliy and Dasia had been working on gained sudden interest and was passed in record time.

When I spoke to Vitaliy several years later, in 2018, the legislation was still in place, but the backsliding had already begun. Undeterred, he told me that you always have to be preparing for your next window of opportunity. You can’t dictate your luck, nor can you predict when it will come, but what you can do is prepare for it.

The Power Of Existential Rebellion

The French writer Albert Camus believed our existence was absurd. He compared the human condition to Sisyphus, the mythical Greek king condemned to roll a boulder uphill, only to see it roll back down, for eternity. Incredibly, Camus imagines Sisyphus, returning to his labors at the foot of the mountain, as happy, having found meaning in his task.

That is the nature of existential rebellion, to find meaning for yourself in a universe that provides none. In two decades researching innovation, transformation and change, one constant I have found is that you can’t control your luck. Anything can happen. “Sure things” often fail while low-probability events occur all the time.

Bill Haley performed “Rock Around The Clock,” because it spoke to him, even over the objections of the record labels. He had no way of knowing it would be a hit for the ages. In a similar way, Einstein pursued physics as a clerk at the Swiss patent office to answer his own questions. Anti-corruption activists worked for years in Ukraine—at great risk to themselves—when it seemed pointless or even, absurd.

Yet it is not hard to imagine Haley joyfully jamming away, even if incredible fortune had not smiled on him, and that Einstein would have lived a fulfilling life even if his miracle year had never happened. Activists like Dasia Kaleniuk and Vitaliy Shabunin continue to investigate corruption in Ukraine, even while being subjected to vicious attacks.

It is a simple truth that we can’t control our luck and luck greatly influences our successes and failures. But we can pursue meaning in things that we define ourselves—an idea, family, justice, compassion or anything else. Or, as the mathematician G.H. Hardy put it, “The case for my life, then… is this: that I have added something…”

— Article courtesy of the Digital Tonto blog
— Image credit: Pexels

Subscribe to Human-Centered Change & Innovation WeeklySign up here to join 17,000+ leaders getting Human-Centered Change & Innovation Weekly delivered to their inbox every week.

Listening to a Ghost

What Miles Davis teaches us about the art of innovation

GUEST POST from John Bessant

Summertime – and, as the song says, the living is easy. Relaxing, recharging mental batteries and absorbing plenty of sunshine. Moments of golden pleasure.

One of which for me is the joy of listening to the ‘Proms’. The famous series of concerts started a century ago by Henry Wood and Robert Newman which aimed to bring serious music to the ears of everyone. Their vision was the ‘promenade’ style of concert, one which allowed people to stand, walk around, and enjoy the music casually instead of sitting in formal, quiet rows. For a long time this wonderful experience was confined to the Royal Albert Hall in London, but when the BBC took over running these events it opened up the concerts to the world via the magic of broadcasting. Democratizing and diverse, bringing all sorts of music to the attention of millions.

Royal Albert Hall

Including people like me, sitting and watching the sunset with a soundtrack. And right now I’m listening to a particularly appropriate piece to accompany Helios as he sinks slowly behind gold-topped hills. A young trumpeter working his skillful way around Gil Evans’ famous arrangement of George Gershwin’s haunting song ‘Summertime’ from the musical Porgy and Bess.

Closing my eyes – and getting shivers down my spine.

Which is an entirely appropriate if you’ve just seen (or in my case, heard) a ghost.

The specter in question would have celebrated his 100th birthday this year and this Promenade concert is being staged to remind us of the huge contribution he made as a musical innovator.

Miles Davis

Miles Davis, born 26th May, 1926; by the time he died he’d rewritten many of the rules of the musical game. Someone whose spirit lives on in the style and approach taken by artists around the world today.

What he knew about making something new – innovation – hasn’t dated at all.

It’s easy to see him as the stereotype of the jazz musician – leaning against a wall, sunglasses worn inside, trumpet dangling loosely from his fingers. Reflecting and radiating an idea which he’d done so much to create – the birth of the cool. The moment when the frenzied and sometimes chaotic energy of jazz moved into a calmer, more meditative kind of exploration. Trying out new combinations of instruments, playing with textures, making use of space.

But it’s what’s behind the image, the way in which he actually worked, which carries the real impact. Because there was undoubted method in his madness, a pattern which we can see recurring from his earliest days.

And it’s a pattern which still has a lot to offer our thinking about innovation

Here are five ‘inspirations’ ….

1. The power of improvisation – create out of what isn’t there, reach for the unknown.

Improvisation

Paris, 1957, coming up on Christmas and Miles is about to go into the recording studio with four musicians with whom he’s not played before. His famous quintet couldn’t make the tour due to other commitments so the hastily assembled quartet are musically competent but comparative strangers. The session is to record film music for a young French director, Louis Malle who has just finished shooting ‘L’ascenseur sur l’echafaud‘ (Elevator to the scaffold). He’s a young man in a hurry, wanting to submit the film for a competition the following autumn. Miles has never made film music before but has agreed to give it a shot; he’s only seen the film once yet the new band are surprised by how he’s calmly approaching the session as if he knows exactly what to do.

Their challenge was the classic blank canvas, being tasked to find something that wasn’t there. Louis Malle explicitly told them not to try and interpret what they saw on the screen but instead try to find a counterpoint to the images. They never saw the film in total, only short excerpts of twenty to thirty seconds on which Davis encouraged them to improvise freely, giving them a minimum of guidance or structure. It was all about atmosphere.

The whole session lasted less than four hours; from it fifty minutes of music emerged, eighteen of which was used for the film. But they were enough; the film went on to win the Prix Louis-Delluc and the soundtrack received its own award the following year, the prestigious Grand Prix du Disc of the Académie Charles-Cros.


It’s good practice for what comes next. By now Miles has refined and developed this style into a trademark approach. Two years on and in a studio in New York a different group of musicians, a sextet this time, are assembled for another recording session. They’re a mixed group in terms of styles; pianist Bill Evans, John Coltrane and Julian (Cannonball) Adderley on saxophones, Paul Chambers on bass and drummer Jimmy Cobb. (Pianist Wynton Kelly joined to replace Evans on one track). Many of the same dynamics were involved; little instruction or preparation beyond the basic challenge of improvisation. Bill Evans described the process in the liner notes on the album sleeve; Davis gave them a few brief instructions and minimal sketches of scales and melody lines on which to improvise.

Recorded at Columbia’s 30th Street Studio in two sessions on March 2 and April 22, 1959, ‘Kind of blue’ is regarded as possibly the greatest jazz album of all time. It is certainly one of the best-known and had enormous influence; it was one of fifty recordings chosen in 2002 by the Library of Congress for the inaugural year of the National Recording Registry, being deemed “culturally, historically, or aesthetically significant”.

The waves it created spread out widely over time and space; many rock musicians directly acknowledge its impact on their thinking in bands as diverse as the Allman Brothers, Pink Floyd, Radiohead, the Velvet Underground, Brian Eno, Steely Dan and the Red Hot Chilli Peppers.

What lay – and still lies – behind the approach which they took was much more than individuals playing around with different notes. One way to describe it might be ‘enabled exploration’ – the chance to find and follow ideas which emerge, discovering places you didn’t know were there, confident in the rest of the band supporting you on that journey. And in turn that depends on two key elements – active listening, being acutely aware of what you’re hearing and where the music seems to be pointing next. And a ‘yes- and…’ mindset which accepts the direction and goes with the flow.

Actors and comedians would recognize this – it’s at the heart of ‘improv’ theater and comedy, a process not just of making it up as you go along but doing so in a way which builds on high levels of trust and psychological safety. And it emerges in countless stories of how radical innovations emerge, by having the space and support to explore in unexpected new ways.

Miles is almost as famous for his quotable sayings as his music and one phrase captures this improvisation approach well.

‘You don’t make bad notes… the note next to the one you think is bad corrects the one in front….’

Herbie Hancock, who joined Davis’s band as a young pianist had direct experience of this early in his life with the band. He played a chord which didn’t work, ‘hanging there like a rotten piece of fruit’ as he put it, and he waited for the inevitable crash which would follow. But Miles didn’t stop; his response was to pick up on what had just happened and then create a run of notes which somehow bent around the ‘wrong ‘chord to make it work. Hancock says it took him years to really absorb this lesson – that there’s no such thing as a wrong note in isolation, only a wrong note nobody responds to.

There’s a lesson there for any innovator and it speaks to more than just platitudes about embracing failure as opportunity. It suggests the discipline and openness needed to bring that to life.

2. Creativity loves constraints

Part of the magic behind Miles’s approach lay not only in working within constraints, like time pressures, availability of musicians and unhelpful physical stages. He’d actively embrace them and seek to restrict his band still further, squeezing them into uncomfortable situations – like the deliberate lack of instruction or preparation behind the studio work for the above projects. This drew on a deep sense that constraints and pressure can sometimes force surprising directions and creative responses; the most unpropitious circumstances can act as a trigger for innovation.

He understood something which runs through so many accounts of surprising flashes of inspired creativity in the least promising situations. Not least the experience of pianist Keith Jarrett, arriving late afternoon to prepare for a concert at the opera house in Cologne. He was tired after a long drive, feeling feverish and would happily have cancelled the solo performance he was contracted to deliver. What he found on the stage was the perfect excuse – instead of the high-quality concert piano he’d specified a mix-up had meant that he was supposed to perform on a rehearsal instrument in need of some major maintenance. This one had pedals which stuck, a weak bass end and a thin top register. And it was out of tune.

It was only the impassioned pleading from the teenage promoter, Vera Brandes, that persuaded him to go on stage and do the show. A good job that she did; Jarrett somehow built an inspired performance out of the very constraints which the instrument threw at him. It was his trademark improvisation but cranked up to build on the flaws, creating driving, percussive left-hand figures standing in for the sustain and depth the instrument didn’t have. The recording, The Köln Concert, became the best-selling solo piano album, and the best-selling solo jazz album, in history.

Others have used this same approach in deliberate fashion. Producer Brian Eno collaborated with the painter Peter Schmidt in the mid-1970s to develop a set of cards which they called Oblique Strategies. Each was printed with an instruction, vaguely phrased but designed to introduce a constraint and force a different action or style. Things like honor thy error as a hidden intention, use an unacceptable color. Every time a recording session became stuck they would draw a card at random. The cards were designed to force a different approach, view the challenge form a different angle.

Film maker Lars von Trier used something similar in his film ‘The five obstructions’ in 2003, made with his friend and mentor, Jørgen Leth. Von Trier’s favorite film is ‘The Perfect Human’ by Leth and in this film he deliberately challenges Leth to remake the film five times, each with a different “obstruction” (or obstacle) imposed by Trier. The resulting film was voted one of the 30 best films of the 2000s in a poll for ‘Sight and Sound’ magazine.


3. Diversity makes a difference

Diversity

By its nature jazz has always been a mixture; its origins are particularly associated with New Orleans where the ‘cradle of jazz’ was home to a vibrant mix of cultures. Jazz combines elements of blues, ragtime, spirituals, work songs, folk traditions, and marching band music and its different styles cover swing rhythms, blue notes, polyrhythms and call-and-response patterns. And it is delivered by individuals converging for temporary collective work before moving on and regrouping elsewhere.

Viewed through this lens the band Miles put together for the ‘Kind of Blue’ sessions was typically diverse – and deliberately chosen to be so. The aim was for creative friction rather than smooth fit. The range was wide, from Bill Evans’s classical harmonic piano to Cannonball Adderley’s blues-infused gospel, with John Coltrane’s tenor pushing against both of them. It was precisely their diversity of styles that allowed the improvisational space to be filled with so many fresh ideas.

It was an idea Miles knew from his early days in the basement apartment where he and Gil Evans rehearsed the tracks which when compiled would yield albums like ‘The birth of the cool’ and ‘Sketches of Spain’. These arrangements featured an unusual nine-piece ensemble (nonet) that included a French horn and a tuba, creating rich, classical-leaning brass textures. Aiming to break away from the ‘heat’ of the bebop big band era and explore something different.

And it was something he was doing decades later, recruiting funk and soul, rock and other craftsmen to his different ensembles. It was a deliberate choice, constantly trying to refresh and challenge; for example the quintet which created key albums like ‘Bitches’ Brew’ was built from a fusion of older experience players with the fresh angle from a drummer, Tony Williams, who was only seventeen when he joined the band.

It’s an approach which has served other artists well; Paul Simon’s reinvention of his style to embrace African rhythms and melodies owes a great deal to his deliberate recruitment of local musicians, in particular Ladysmith Black Mambazo.

Much of the power of diversity has its roots in the idea of the ‘jam session’, an informal and often after-hours coming together of players who don’t normally play together but who exchange ideas and tricks. Like the night back in 1966 when a 23 year old guitarist asked if he could sit in with Cream during a gig at the Regent Street Polytechnic. An intrigued Eric Clapton agreed and was treated to a version of Howlin’ Wolf’s “Killing Floor” played at massively accelerated speed and accompanied by a whole repertoire of flashy techniques. It prompted him to say to Chas Chandler, the manager of the then-unknown Jimi Hendrix ‘you never told me he was that f***ing good!’

This kind of jamming isn’t unusual; it’s widely recognized as a crucible from which innovation can occur. But it depends on ‘acceptable accidents’ – an approach which allows the space for it to occur. Embracing uncertainty, allowing difference to create.


4. Reinvention – deliberate (creative) destruction

A look back across half a century of making music reveals another aspect of Miles’ creative approach – his deliberate dismantling of groups while they were often still at the height of their popularity. With almost rhythmic precision he’d take apart a sound which critics and audiences had come to love every five years or so. His changes – from the frenzied bebop sounds with which he began, through the cool jazz of the fifties, to the electric rock sprawl of Bitches Brew and beyond – ran like waves of ‘creative destruction’ across his music. It wasn’t a case of him running out of ideas; it was a constant search for new space to explore.

Again he’s in good company with such an approach. Picasso did the same thing across painting, his phases now neatly labelled in galleries and museums. Blue, Rose, Cubist, Neoclassical, and so on. David Bowie did it within a single decade, discarding Ziggy Stardust to become the Thin White Duke, then jettisoning that persona as he moved to Berlin to make electronic records with Eno.

Reinvention of this kind isn’t a response to a failing output; instead it’s about riding the waves of success and then deliberately choosing to jump away to retain the element of surprise.


5. Space as an enabling structure

So much of what Miles practiced links back to the idea of space – allowing somewhere for ideas to emerge and be explored. This was clear in his instructions to musicians, reminding them that ‘the most important part of music is the silence between the notes’.

A concept not lost on John Coltrane; early on he complained that he didn’t know how to end his solos to which Miles drily responded ‘take the horn out of your mouth, John!’

But the underlying concept mattered; Coltrane’s own impressive career repeatedly references the way in which he learned to find the space within which he could create his famous sheets of sound.

It’s easy to see this as just another reflection on improvisation which builds on active and attentive listening. But it’s also something which Miles learned and refined in his early work in Gil Evans’s basement, working with large ensembles. A piece like the haunting version of the ‘Concierto d’Aranjuez’ on the ‘Sketches of Spain’ album doesn’t emerge from letting musicians find their own way; it depends on careful orchestration to simultaneously allow structure and carefully sheltered space within which improvisation becomes possible.

As he famously explained his memorable treatment came to him by allowing the melody to occupy the space Evans had created; ‘….. so strong that the softer you play it, the stronger it gets, and the stronger you play it, the weaker it gets’.

Working in this way poses challenges very much like those of a large organization trying to retain an entrepreneurial heart inside a well-oiled and structured machine. Too much improvisation space – and there’s a risk of serious disruption; too little and the drift to sterility emerges. It’s a delicate balancing act – the kind of ambidexterity which organization theorists often talk about.


Αnd it’s that piece – the lovely adagio with its breathy tones, the fluttering of castanets and the shimmering of strings like desert heat rising from the Spanish plains – which I’m currently enjoying as a perfect accompaniment to my sunset.

A great way to rekindle some innovation inspiration.


All images public domain or generated by Gemini AI unless otherwise indicated

You can listen to a podcast of this story here

And watch a video version on my YouTube channel here

If you’d like more songs, stories and other resources on the innovation theme, check out my website here

And if you’d like to learn with me take a look at my online courses here

Subscribe to Human-Centered Change & Innovation WeeklySign up here to join 17,000+ leaders getting Human-Centered Change & Innovation Weekly delivered to their inbox every week.

What Others Are Saying About the Importance of Experience Audits

What Others Are Saying About the Importance of Experience Audits

by Braden Kelley and Art Inteligencia


I have an obvious stake in this topic, so let me say it up front: I offer experience audits, and I think they’re valuable. That makes me a poor judge of whether you should believe me.

So I went looking for what other people say. Consultants, researchers, a Big Four firm, a government order, and companies that sell software to channel teams. They work in different industries, look at different audiences and, in most cases, have never heard of each other. And yet when I lined their words up, the same handful of ideas kept showing up.

Here’s what I found, organized by the kind of audit each person was talking about: customer, employee, partner and citizen.

Customer experience audits: seeing what you’ve stopped seeing

The customer side is the best-documented, so let’s start there. The recurring theme is that an audit gives you something your internal reports can’t: an honest, outside view.

Customer experience audit

“An audit provides value to an organization through an independent view of all the customer touch points.”

Bryan Comite, managing director at Protiviti, writing in Quirk’s, June 2018

The word doing the work there is independent. People inside an organization have walked the same path so many times that the friction has become invisible to them. Someone from outside sees it in the first ten minutes.

Customer experience audit

“A thorough audit uncovers the real, sometimes gritty truth of how customers interact with your brand at every touchpoint.”

Lauren Basiura, Marketing Insider Group, May 2024

I like “gritty” there, because it’s accurate. The truth about your customer experience is rarely flattering and rarely tidy. It lives in the confusing confirmation email and the phone tree that loses people, not in a dashboard.

Customer experience audit

“By putting yourself in your customers’ shoes during an audit, you’ll naturally stumble into customer pain points.”

Ryan Kane, The CX Lead, updated February 2026

Stumble into is exactly right. You don’t need a sophisticated analytical model to find the biggest problems. You need to experience the journey the way a customer does, which is why walking it firsthand sits at the center of any audit worth doing.

Customer experience audit

“Customer experience audits offer something every organisation needs: clarity.”

Kaarle Parikka, Falcony, April 2026

Clarity is the word I’d put on the side of the building. Most organizations don’t lack data. They lack a clear picture of what the data means.

Employee experience audits: the score isn’t the story

On the employee side, the most useful observation I found was about the limits of the number most organizations rely on.

Employee experience audit

“An engagement score tells you the temperature. It rarely tells you which moment in the journey set it.”

Mutomorro, employee experience audit service page

Every leader who has ever stared at a flat or falling engagement score knows this feeling. You know something is wrong. You don’t know where. The same page makes the next point that matters:

Employee experience audit

“The audit finds that moment, so the fix is to redesign it rather than nudge the number.”

Mutomorro, same page

Redesign it rather than nudge the number is a useful warning against the most common mistake in this area, which is launching a program aimed at the score instead of at the experience that produces it.

Employee experience audit

“The goal of this audit is clarity, not complexity.”

Advantage Club, January 2026

Clarity again. It’s becoming a theme.

Partner experience audits: the experience you don’t own, but customers blame you for

Partner experience gets far less attention than either customers or employees, and I suspect that’s why it’s where many organizations are most exposed. Your resellers, distributors, integrators and suppliers deliver pieces of your customer experience, and when they struggle to work with you, your customers feel it.

Partner experience audit

“If you can’t see how your partners actually feel, you can’t fix what’s broken.”

Marketopia, June 2026

The same article makes an observation about how most vendors check in with partners: “Most vendor channel programs measure partner satisfaction the same way. Anecdotally.” Anecdotes are fine for color. They’re a poor basis for investment.

Partner experience audit

“As noted in research from PartnerPath, nearly 70% of vendors do not measure partner experience.”

Trevor Burnett, Impartner, October 2023

Read that again. Nearly 70% don’t measure it at all. If that’s even roughly right, then a partner experience audit is the rare opportunity to learn something your competitors probably don’t know either.

Partner experience audit

“Collecting partner feedback without acting on it is worse than not collecting it at all.”

Unifyr, Partner Experience glossary entry

I’d put that on a poster in every department that runs surveys, and not only in the channel team.

Citizen experience audits: measuring dignity, not just delivery

Governments face the same problem as companies, with higher stakes and fewer alternatives for the people on the receiving end. The best line I found on the public-sector side was also the most uncomfortable.

Citizen experience audit

“Most governments measure whether a service was delivered. Very few measure whether the person receiving it felt treated with dignity.”

Julian Ford, Renascence, August 2026

Delivered versus dignified is a gap that applies to far more than government, and I’d invite any organization to ask which side of it they’re measuring.

Citizen experience audit

“When improving citizen experience, government organisations need to get to the root of what citizens really need.”

KPMG UK, Re-visit citizen journeys, July 2023

The root, not the symptom. And at the policy level, the U.S. federal government made a version of the same argument in a 2021 order on customer experience, which called for services “driven fundamentally by the voice of the customer” and pointed agencies to “human-centered design methodologies such as journey mapping.”

What they all have in common

Put these voices side by side and a pattern emerges, regardless of whether the audience is a customer, an employee, a partner or a citizen.

Scores and surveys can’t tell you where the problem is. Whether it’s an engagement score, a satisfaction rating or the “anecdotal” feedback on a channel program, the measurement tells you something is off without telling you what or where.

An outside view sees what insiders can’t. The word “independent” appears in the customer quotes, and the same idea runs through the others. If you’ve lived with a process for years, you’ve stopped noticing it.

The point is clarity, then action. Two different writers reach for the same word. A clear picture is the goal, and a clear picture that nobody acts on is the failure mode the partner quote warns about.

You have to walk it. “Put yourself in your customers’ shoes.” “Map the journey from the citizen’s starting point, not the department’s.” The findings that matter most come from experiencing the journey, not reading about it.

A fair caution

I should be honest about the limits of this exercise. Many of the voices above work for companies that sell related services or software, so they have a reason to agree with me. Their view isn’t neutral, and neither is mine. What I’d ask you to notice is that they arrived at similar conclusions from very different starting points, and that the conclusions hold up against your own experience. Think about the last time a score told you something was wrong without telling you why.

Where to start

If you’re not sure which audience to look at first, I’d start where your revenue or your risk is concentrated, and add the others as you can. If you’d like to put a number on what friction is costing you before you commit to anything, the CX ROI Calculator and the Employee Experience ROI Calculator are both free. If a few terms in this article are new to your team, the Experience Design Glossary will get everyone speaking the same language.

And if you’d like to talk about what an independent experience audit would look like for your organization, the Customer Experience Audit page explains how it works, and I’m glad to talk through your situation.

Sources

Image Credits: ChatGPT

Content Authenticity Statement: The topic area, key elements to focus on, etc. were decisions made by Braden Kelley, with a little help from Claude to clean up the article.

Subscribe to Human-Centered Change & Innovation WeeklySign up here to get Human-Centered Change & Innovation Weekly delivered to your inbox every week.