Category Archives: Design

Cutting-Edge Ways to Decouple Data Growth from Power and Water Consumption

The Sustainability Imperative

LAST UPDATED: November 1, 2025 at 8:59 AM

Cutting-Edge Ways to Decouple Data Growth from Power and Water Consumption

GUEST POST from Art Inteligencia

The global digital economy runs on data, and data runs on power and water. As AI and machine learning rapidly accelerate our reliance on high-density compute, the energy and environmental footprint of data centers has become an existential challenge. This isn’t just an engineering problem; it’s a Human-Centered Change imperative. We cannot build a sustainable future on an unsustainable infrastructure. Leaders must pivot from viewing green metrics as mere compliance to seeing them as the ultimate measure of true operational innovation — the critical fuel for your Innovation Bonfire.

The single greatest drain on resources in any data center is cooling, often accounting for 30% to 50% of total energy use, and requiring massive volumes of water for evaporative systems. The cutting edge of sustainable data center design is focused on two complementary strategies: moving the cooling load outside the traditional data center envelope and radically reducing the energy consumed at the chip level. This fusion of architectural and silicon-level innovation is what will decouple data growth from environmental impact.

The Radical Shift: Immersive and Locational Cooling

Traditional air conditioning is inefficient and water-intensive. The next generation of data centers is moving toward direct-contact cooling systems that use non-conductive liquids or leverage natural environments.

Immersion Cooling: Direct-to-Chip Efficiency

Immersion Cooling involves submerging servers directly into a tank of dielectric (non-conductive) fluid. This is up to 1,000 times more efficient at transferring heat than air. There are two primary approaches: single-phase (fluid remains liquid, circulating to a heat exchanger) and two-phase (fluid boils off the server, condenses, and drips back down).

This method drastically reduces cooling energy and virtually eliminates water consumption, leading to Power Usage Effectiveness (PUE) ratios approaching the ideal 1.05. Furthermore, the fluid maintains a more stable, higher operating temperature, making the waste heat easier to capture and reuse, which leads us to our first case study.

Case Study 1: China’s Undersea Data Center – Harnessing the Blue Economy

China’s deployment of a commercial Undersea Data Center (UDC) off the coast of Shanghai is perhaps the most audacious example of locational cooling. This project, developed by Highlander and supported by state entities, involves submerging sealed server modules onto the seabed, where the stable, low temperature of the ocean water is used as a natural, massive heat sink.

The energy benefits are staggering: developers claim UDCs can reduce electricity consumption for cooling by up to 90% compared to traditional land-based facilities. The accompanying Power Usage Effectiveness (PUE) target is below 1.15 — a world-class benchmark. Crucially, by operating in a closed system, it eliminates the need for freshwater entirely. The UDC also draws nearly all its remaining power from nearby offshore wind farms, making it a near-zero carbon, near-zero water compute center. This bold move leverages the natural environment as a strategic asset, turning a logistical challenge (cooling) into a competitive advantage.

Case Study 2: The Heat Reuse Revolution at a Major Cloud Provider

Another powerful innovation is the shift from waste heat rejection to heat reuse. This is where true circular economy thinking enters data center design. A major cloud provider (Microsoft, with its various projects) has pioneered systems that capture the heat expelled from liquid-cooled servers and redirect it to local grids.

In one of their Nordic facilities, the waste heat recovered from the servers is fed directly into a local district heating system. The data center effectively acts as a boiler for the surrounding community, warming homes, offices, and water. This dramatically changes the entire PUE calculation. By utilizing the heat rather than simply venting it, the effective PUE dips well below the reported operational figure, transforming the data center from an energy consumer into an energy contributor. This demonstrates that the true goal is not just to lower consumption, but to create a symbiotic relationship where the output of one system (waste heat) becomes the valuable input for another (community heating).

“The most sustainable data center is the one that gives back more value to the community than it takes resources from the planet. This requires a shift from efficiency thinking to regenerative design.”

Innovators Driving the Sustainability Stack

Innovation is happening at every layer, from infrastructure to silicon:

Leading companies and startups are rapidly advancing sustainable data centers. In the cooling space, companies like Submer Technologies specialize in immersion cooling solutions, making it commercially viable for enterprises. Meanwhile, the power consumption challenge is being tackled at the chip level. AI chip startups like Cerebras Systems and Groq are designing new architectures (wafer-scale and Tensor Streaming Processors, respectively) that aim to deliver performance with vastly improved energy efficiency for AI workloads compared to general-purpose GPUs. Furthermore, cloud infrastructure provider Crusoe focuses on powering AI data centers exclusively with renewable or otherwise stranded, environmentally aligned power sources, such as converting flared natural gas into electricity for compute, tackling the emissions challenge head-on.

The Future of Decoupling Growth

To lead effectively in the next decade, organizations must recognize that the convergence of these technologies — immersion cooling, locational strategy, chip efficiency, and renewable power integration — is non-negotiable. Data center sustainability is the new frontier for strategic change. It requires empowered agency at the engineering level, allowing teams to move fast on Minimum Viable Actions (MVAs) — small, rapid tests of new cooling fluids or localized heat reuse concepts — without waiting for monolithic, years-long CapEx approval. By embedding sustainability into the very definition of performance, we don’t just reduce a footprint; we create a platform for perpetual, human-driven innovation.

You can learn more about how the industry is adapting to these challenges in the face of rising heat from AI in the video:

This video discusses the limitations of traditional cooling methods and the necessity of liquid cooling solutions for next-generation AI data centers.

Disclaimer: This article speculates on the potential future applications of cutting-edge scientific research. While based on current scientific understanding, the practical realization of these concepts may vary in timeline and feasibility and are subject to ongoing research and development.

UPDATE: Apparently, Microsoft has been experimenting with underwater data centers for years and you can learn more about them and progress in this area in this video here:

Image credit: Google Gemini

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

Picking Innovation Projects in Four Questions or Less

Leave a reply

Picking Innovation Projects in Four Questions or Less

GUEST POST from Mike Shipulski

It’s a challenge to prioritize and choose innovation projects. There are open questions on the technology, the product/service, the customer, the price and sales volume. Other than that, things are pretty well defined.

But with all that, you’ve still go to choose. Here are four questions that may help in your selection process:

1. Is it big enough?

The project will be long, expensive and difficult. And if the potential increase in sales is not big enough, the project is not worth starting. Think (Price – Cost) x Volume. Define a minimum viable increase in sales and bound it in time. For example, the minimum incremental sales is twenty five million dollars after five years in the market. If the project does not have the potential to meet those criteria, don’t do the project. The difficult question – How to estimate the incremental sales five years after launch? The difficult answer – Use your best judgement to estimate sales based on market size and review your assumptions and predictions with seasoned people you trust.

2. Why you?

High growth markets/applications are attractive to everyone, including the big players and the well-funded start-ups. How does your company have an advantage over these tough competitors? What about your company sets you apart? Why will customers buy from you? If you don’t have good answers, don’t start the project. Instead, hold the work hostage and take the time to come up with good answers. If you come up with good answers, try to answer the next questions. If you don’t, choose another project.

3. How is it different?

If the new technology can’t distinguish itself over existing alternatives, you don’t have a project worth starting. So, how is your new offering (the one you’re thinking about creating) better than the ones that can be purchased today? What’s the new value to the customer? Or, in the lingo of the day, what is the Distinctive Value Proposition (DVP)? If there’s no DVP, there’s no project. If you’re not sure of the DVP, figure that out before investing in the project. If you have a DVP but aren’t sure it’s good enough, figure out how to test the DVP before bringing the DVP to life.

4. Is it possible?

Usually, this is where everyone starts. But I’ve listed it last, and it seems backward. Would you rather spend a year making it work only to learn no one wants it, or would you rather spend a month to learn the market wants it then a year making it work? If you make it work and no one wants it, you’ve wasted a year. If, before you make it work, you learn no one wants it, you’ve spent a month learning the right thing and you haven’t spent a year working on the wrong thing. It feels unnatural to define the market need before making it work, but though it feels unnatural, it can block resources from working on the wrong projects.

Conclusion

There is no foolproof way to choose the best innovation projects, but these four questions go a long way. Create a one-page template with four sections to ask the questions and capture the answers. The sections without answers define the next work. Define the learning objectives and the learning activities and do the learning. Fill in the missing answers and you’re ready to compare one project to another.

Sort the projects large-to-small by Is it big enough? Then, rank the top three by Why you? and How is it different? Then, for the highest ranked project, do the work to answer Is it possible?

If it’s possible, commercialize. If it’s not, re-sort the remaining projects by Is it big enough? Why you? and How is it different? and learn if It is possible.

HALLOWEEN BONUS: Save 30% on the eBook, hardcover or softcover of Braden Kelley’s latest book Charting Change (now in its second edition) — FREE SHIPPING WORLDWIDE — using code HAL30 until midnight October 31, 2025

Image credit: Pexels

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

How Tangible AI Artifacts Accelerate Learning and Alignment

Leave a reply

Seeing the Invisible

By Douglas Ferguson, Founder & CEO of Voltage Control
Originally inspired by “A Lantern in the Fog” on Voltage Control, where teams learn to elevate their ways of working through facilitation mastery and AI-enabled collaboration.

Innovation isn’t just about generating ideas — it’s about testing assumptions before they quietly derail your progress. The faster a team can get something tangible in front of real eyes and minds, the faster they can learn what works, what doesn’t, and why.

Yet many teams stay stuck in abstraction for too long. They debate concepts before they draft them, reason about hypotheses before they visualize them, and lose energy to endless interpretation loops. That’s where AI, when applied strategically, becomes a powerful ally in human-centered innovation — not as a shortcut, but as a clarifier.

How Tangible AI Artifacts Accelerate Learning and Alignment

At Voltage Control, we’ve been experimenting with a practice we call AI Teaming — bringing AI into the collaborative process as a visible, participatory teammate. Using new features in Miro, like AI Flows and Sidekicks, we’re able to layer prompts in sequence so that teams move from research to prototypes in minutes. We call this approach Instant Prototyping — because the prototype isn’t the end goal. It’s the beginning of the real conversation.

Tangibility Fuels Alignment

In human-centered design, the first artifact is often the first alignment. When a team sees a draft — even one that’s flawed — it changes how they think and talk. Suddenly, discussions move from “what if” to “what now.” That’s the tangible magic: the moment ambiguity becomes visible enough to react to.

AI can now accelerate that moment. With one-click flows in Miro, facilitators can generate structured artifacts — such as user flows, screen requirements, or product briefs — based on real research inputs. The output isn’t meant to be perfect; it’s meant to be provocative. A flawed draft surfaces hidden assumptions faster than another round of theorizing ever could.

Each iteration reveals new learning: the missing user story, the poorly defined need, the contradiction in the strategy. These insights aren’t AI’s achievement — they’re the team’s. The AI simply provides a lantern, lighting up the fog so humans can decide where to go next.

Layering Prompts for Better Hypothesis Testing

One of the most powerful aspects of Miro’s new AI Flows is the ability to layer prompts in connected sequences. Instead of a single one-off query, you create a chain of generative steps that build on each other. For example:

Synthesize research into user insights.
Translate insights into “How Might We” statements.
Generate user flows based on selected opportunities.
Draft prototype screens or feature lists.

Each layer of the flow uses the prior outputs as inputs — so when you adjust one, the rest evolves. Change a research insight or tweak your “How Might We” framing, and within seconds, your entire prototype ecosystem updates. It’s an elegant way to make hypothesis testing iterative, dynamic, and evidence-driven.

Seeing the Invisible

In traditional innovation cycles, these transitions can take weeks of hand-offs. With AI flows, they happen in minutes — creating immediate feedback loops that invite teams to think in public and react in real time.

(You can see this process in action in the video embedded below — where we walk through how small prompt adjustments yield dramatically different outputs.)

The Human Element: Facilitating Sensemaking

The irony of AI-assisted innovation is that the faster machines generate, the more valuable human facilitation becomes. Instant prototypes don’t replace discussion — they accelerate it. They make reflection, critique, and sensemaking more productive because there’s something concrete to reference.

Facilitators play a critical role here. Their job is to:

Name the decision up front: “By the end of this session, we’ll have a directionally correct concept we’re ready to test.”
Guide feedback: Ask, “What’s useful? What’s missing? What will we try next?”
Anchor evidence: Trace changes to specific research insights so teams stay grounded.
Enable iteration: Encourage re-running the flow after prompt updates to test the effect of new assumptions.

Through this rhythm of generation, reflection, and adjustment, AI becomes a conversation catalyst — not a black box. And the process stays deeply human-centered because it focuses on learning through doing.

Case in Point: Building “Breakout Buddy”

We recently used this exact approach to prototype a new tool called Breakout Buddy — a Zoom app designed to make virtual breakout rooms easier for facilitators. The problem was well-known in our community: facilitators love the connection of small-group moments but dread the logistics. No drag-and-drop, no dynamic reassignment, no simple timers.

Using our Instant Prototyping flow, we gathered real facilitator pain points, synthesized insights, and created an initial app concept in under two hours. The first draft had errors — it misunderstood terms like “preformatted” and missed saving room configurations — but that’s precisely what made it valuable. Those gaps surfaced the assumptions we hadn’t yet defined.

After two quick iterations, we had a working prototype detailed enough for a designer to polish. Within days, we had a testable artifact, a story grounded in user evidence, and a clear set of next steps. The magic wasn’t in the speed — it was in how visible our thinking became.

Designing for Evidence, Not Perfection

If innovation is about learning, then prototypes are your hypotheses made tangible. AI just helps you create more of them — faster — so you can test, compare, and evolve. But the real discipline lies in how you use them.

Don’t rush past the drafts. Study what’s wrong and why.
Don’t hide your versions. Keep early artifacts visible to trace the evolution.
Don’t over-polish. Each iteration should teach, not impress.

When teams treat AI outputs as living evidence rather than final answers, they stay in the human-centered loop — grounded in empathy, focused on context, and oriented toward shared understanding.

A Lantern in the Fog

At Voltage Control, we see AI not as a replacement for creative process, but as a lantern in the fog — illuminating just enough of the path for teams to take their next confident step. Whether you’re redesigning a product, reimagining a service, or exploring cultural transformation, the goal isn’t to hand creativity over to AI. It’s to use AI to make your learning visible faster.

Because once the team can see it, they can improve it. And that’s where innovation truly begins.

🎥 Watch the Demo: How layered AI prompts accelerate hypothesis testing in Miro

Join the waitlist to get your hands on the Instant Prototyping template

Image Credit: Douglas Ferguson, Unsplash

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

An Industrial Call-To-Arms for the Environment

Leave a reply

An Industrial Call-To-Arms for the Environment

GUEST POST from Mike Shipulski

What is your obligation to improve the health of our planet?

For the CEO – Look around. Look at Europe. Look at China’s plans. Look at the startups. I know you want to achieve your growth objectives, but if you don’t take seriously the race toward cleaner products and services, you’ll go out of business. You can see this as a problem or an opportunity. Bury your head or put on your track shoes and run! It’s your choice.

Look at the oceans. Look at the landfills. Look at the rise in global temperatures. Just look. This isn’t about ROI, this is about survival. Growth objectives aside, no one will buy things when they are struggling to survive in an uncertain future. Your same old dirty products won’t cut it anymore. So, what are you going to do?

For an example of a path forward, look to the companies in the oil business. Their recipe is clear. They’ve got to use their large but ever-diminishing profits to buy themselves into technologies and industries that will ultimately eat their core business. Though the timing is uncertain, it’s certain that improvements in cleaner technologies will demand they make the change.

Whatever you do, don’t wait. You don’t have much time. Cleaner technologies are getting better every day. It’s time to start.

For Marketing – Look at the upstarts. Look at the powerful companies in adjacent markets who will soon be your direct competitors. Look at your stodgy, unprofitable competitors who are now sufficiently desperate to try anything. Their next marketing push will be built on the bedrock of an improved planet. They’ll be almost as good as you in the traditional areas of productivity and quality and they’ll blow your doors off with their meaner and greener products. Customers will choose green over brown. And they’ll look for real improvements that make the planet smile. The time for green-washing is past. That trick is out of gas.

You need to help customers with new jobs to be done. They care about their environment. They care about their carbon footprint. They care about clean water. And they care about recycling and reuse. It’s real. They care. Now it’s up to you to help them make progress in these areas. It will be a tough road to convince your company that things need to change, but that’s why you’re in Marketing.

You’re already behind. It’s time to start. And it’s up to you to lead the charge.

For Manufacturing – Look at your Value Stream Maps (VSMs). Assign a carbon footprint to each link in the chain. And do the same with water consumption. Assess each process step for carbon and water and rank them worst to best. For the worst, run carbon kaizens and improve the carbon footprint. And run water kaizens for the thirstiest processes.

And look again at your VSMs, and look more broadly. Look back into the supply chain, rank for carbon and water and improve the ones that need the treatment. And teach your suppliers how to do it. And look forward into your distribution channels and improve or eliminate the worst actors. And then propose to Marketing that you teach your customers how to use VSMs to clean up their act. And challenge Engineering to change the design to eliminate the remaining bad actors.

You’ve made good progress with your value streams. Now it’s time to help others make the progress that must be made. As subject matter experts, it’s your time to shine. And, please, start now.

For Engineering – Look at your products. Look at how they’re used. Look at how they’re delivered. Look at how they’re made. Look at how they’re recycled. Sure, your products provide good functionality, but throughout their life cycle they also create carbon dioxide and consume water. And you’re the only ones that can design out the environmental impact.

Learn how to do a Life Cycle Assessment (LCA). Learn which elements of the product create the largest problems. For all the parts that make up the product, sort them worst to best to prioritize the design work. It’s time for radical part count reduction. Try to design out half the parts. It’s possible. And the payoff is staggering. What’s the carbon footprint of a part that was designed out of the product?

Or, to make a more radical improvement, consider an Innovation Burst Event (IBE) to make a fundamental change in the way your products/services impact the environment. With this approach, your innovation work, by definition, will make the planet smile.

It’s time to be open-minded. Ask Manufacturing for the worst processes (including supply chain and distribution) and try to design them out. Design out the part, or change the material, or change the design to enable a friendlier process. Manufacturing can only improve a bad process, but you can design them out altogether. There’s power in that, but with power comes responsibility.

And it’s time for you to take responsibility.

For Everyone in Industry – Regardless of your company, your country or your political affiliation, we can all agree that all our lives get better as the health of our planet improves. And everyone can agree that cleaner air is better. And everyone can agree it’s the same for our water – cleaner is better. And that’s a whole lot of agreement.

As industry leaders, I challenge you to build on that common ground. As industry leaders, I challenge you to improve our planet one product at a time and one process at a time. And as industry leaders, I challenge you to help each other. There’s no competitive disadvantage when you help a company outside your industry. And there’s no shame in learning from companies outside your industry. And it’s good for the planet and profits. There’s nothing in the away. It’s time to start.

As an industry leader, if you want to make a difference in the health of our planet, drop a comment.

Image credit: Pixabay

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

My Advice to Young Design Engineers

Leave a reply

My Advice to Young Design Engineers

GUEST POST from Mike Shipulski

If your solution isn’t sold to a customer, you didn’t do your job. Find a friend in Marketing.

If your solution can’t be made by Manufacturing, you didn’t do your job. Find a friend in Manufacturing.

Reuse all you can, then be bold about trying one or two new things.

Broaden your horizons.

Before solving a problem, make sure you’re solving the right one.

Don’t add complexity. Instead, make it easy for your customers.

Learn the difference between renewable and non-renewable resources and learn how to design with the renewable ones.

Learn how to do a Life Cycle Assessment.

Learn to see functional coupling and design it out.

Be afraid but embrace uncertainty.

Learn how to communicate your ideas in simple ways. Jargon is a sign of weakness.

Before you can make sure you’re solving the right problem, you’ve got to know what problem you’re trying to solve.

Learn quickly by defining the tightest learning objective.

Don’t seek credit, seek solutions. Thrive, don’t strive.

Be afraid, and run toward the toughest problems.

Help people. That’s your job.

Image credit: 1 of 950+ FREE quote slides available at <a href=”http://misterinnovation.com” target=”_blank”>http://misterinnovation.com</a>

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

Why Context Engineering is the Next Frontier in AI

Leave a reply

Why Context Engineering is the Next Frontier in AI

by Braden Kelley and Art Inteligencia

Observing the rapid evolution of artificial intelligence, one thing has become abundantly clear: while raw processing power and sophisticated algorithms are crucial, the true key to unlocking AI’s transformative potential lies in its ability to understand and leverage context. We’ve seen remarkable advancements in generative AI and machine learning, but these technologies often stumble when faced with the nuances of real-world situations. This is why I believe context engineering – the discipline of explicitly designing and managing the contextual information available to AI systems – is not just an optimization, but the next fundamental frontier in AI innovation.

Think about human intelligence. Our ability to understand language, make decisions, and solve problems is deeply rooted in our understanding of context. A single word can have multiple meanings depending on the sentence it’s used in. A request can be interpreted differently based on the relationship between the people involved or the situation at hand. For AI to truly augment human capabilities and integrate seamlessly into our lives, it needs a similar level of contextual awareness. Current AI models often operate on relatively narrow inputs, lacking the broader understanding of user intent, environmental factors, and historical interactions that humans take for granted. Context engineering aims to bridge this gap, moving AI from being a powerful but often brittle tool to a truly intelligent and adaptable partner.

In the realm of artificial intelligence, context engineering is the strategic and human-centered practice of providing an AI system with the relevant background information it needs to understand a query or situation accurately. It goes beyond simple prompt design by actively building and managing the comprehensive context that surrounds an interaction. This includes integrating historical data, user profiles, real-time environmental factors, and external knowledge sources, allowing the AI to move from a narrow, transactional understanding to a more holistic, human-like awareness. By engineering this context, we enable AI to produce more accurate, personalized, and genuinely useful responses, bridging the gap between a machine’s logic and the nuanced complexity of human communication and problem-solving.

The field of context engineering encompasses a range of techniques and strategies focused on providing AI systems with relevant and actionable context. This includes:

Prompt Engineering: Crafting detailed and context-rich prompts that guide AI models towards desired outputs.
Memory Management: Implementing mechanisms for AI to remember past interactions and use that history to inform current responses.
External Knowledge Integration: Connecting AI systems to external databases, APIs, and real-time data streams to provide up-to-date and relevant information.
User Profiling and Personalization: Leveraging data about individual users to tailor AI responses to their specific needs and preferences.
Situational Awareness: Incorporating real-world contextual cues, such as location, time of day, and user activity, to make AI more responsive to the current situation.

A Human-Centered Blueprint for Implementation

Implementing context engineering is not a one-time technical fix; it is a continuous, human-centered practice that must be embedded into your innovation lifecycle. To move beyond a static, one-size-fits-all model and create truly intelligent, context-aware AI, consider this blueprint for action:

Step 1: Start with the Human Context. Before you even think about data streams or algorithms, you must first deeply understand the human being you are serving. Conduct ethnographic research, user interviews, and journey mapping to identify what context is truly relevant to your users. What are their goals? What unspoken needs do they have? What external factors influence their decisions? The most valuable context often isn’t in a database—it’s in the real-world experiences and emotional states of your users.
Step 2: Map the Contextual Landscape. Once you understand the human context, you can begin to identify and integrate the necessary data. This involves creating a “contextual map” that connects the human need to the available data sources. For a customer service AI, this map would link a customer’s inquiry to their purchase history, recent support tickets, and even their browsing behavior on your website. For a medical AI, the map would link a patient’s symptoms to their genetic data, environmental exposure, and family medical history. This mapping process ensures that the AI’s inputs are directly tied to what matters most to the user.
Step 3: Build a Dynamic Feedback Loop. The context of a situation is constantly changing. A great context-aware AI is not a static system but a learning one. Implement a continuous feedback loop where human users can correct the AI’s understanding, provide additional information, and refine its responses. This “human-in-the-loop” approach is vital for ethical and accurate AI. It allows the system to learn from its mistakes and adapt to new, unforeseen contexts, ensuring its relevance and reliability over time.
Step 4: Prioritize Privacy and Ethical Guardrails. The more context you provide to an AI, the more critical it becomes to manage that information responsibly. From the outset, you must design for privacy, collecting only the data you absolutely need and ensuring it is stored and used in a secure and transparent manner. Establish clear ethical guardrails for how the AI uses and interprets contextual information, particularly for sensitive data. This is not just a regulatory requirement; it is a fundamental aspect of building trust with your users and ensuring that your AI serves humanity, rather than exploiting it.

By following these best practices, you can move beyond simple, reactive AI to a proactive, human-centered intelligence that understands the world not just as a collection of data points, but as a rich tapestry of interconnected context. This is the work that will define the next generation of AI and, in doing so, will fundamentally change how technology serves humanity.

Case Study 1: Improving Customer Service with Context-Aware AI Assistants

The Challenge: Generic and Frustrating Customer Service Chatbots

Many companies have implemented AI-powered chatbots to handle customer inquiries. However, these chatbots often struggle with complex or nuanced issues, leading to frustrating experiences for customers who have to repeat information or are given irrelevant answers. The lack of contextual awareness is a major limitation.

Context Engineering in Action:

A telecommunications company sought to improve its customer service chatbot by implementing robust context engineering. They integrated the chatbot with their CRM system, allowing it to access the customer’s purchase history, past interactions, and current account status. They also implemented memory management so the chatbot could retain information shared earlier in the conversation. Furthermore, they used prompt engineering to guide the chatbot to ask clarifying questions and to tailor its responses based on the specific product or service the customer was inquiring about. For example, if a customer asked about a billing issue, the chatbot could access their latest bill and provide specific details, rather than generic troubleshooting steps. It could also remember if the customer had contacted support recently for a related issue and take that into account.

The Impact:

The context-aware chatbot significantly improved customer satisfaction scores and reduced the number of inquiries that had to be escalated to human agents. Customers felt more understood and received more relevant and efficient support. The company also saw a decrease in customer churn. This case study highlights how context engineering can transform a basic AI tool into a valuable and helpful resource by enabling it to understand the customer’s individual situation and history.

Key Insight: By providing AI customer service assistants with access to relevant customer data and interaction history, companies can significantly enhance the quality and efficiency of support, leading to increased customer satisfaction and loyalty.

Case Study 2: Enhancing Medical Diagnosis with Contextual Patient Information

The Challenge: Over-reliance on Isolated Symptoms in AI Diagnostic Tools

AI is increasingly being used to assist medical professionals in diagnosing diseases. However, early AI diagnostic tools often focused primarily on analyzing individual symptoms in isolation, potentially missing crucial contextual information such as the patient’s medical history, lifestyle, environmental factors, and even subtle cues from their recent health records.

Context Engineering in Action:

A research hospital in the Pacific Northwest developed an AI-powered diagnostic tool for a specific type of rare disease. Recognizing the importance of context, they engineered the AI to integrate a wide range of patient data beyond just the presenting symptoms. This included the patient’s complete medical history (past illnesses, medications, allergies), family medical history, lifestyle information (diet, exercise, smoking habits), recent lab results, and even notes from previous doctor’s visits. The AI was also connected to relevant medical literature to understand the broader context of the disease and potential co-morbidities. By providing the AI with this rich contextual information, the researchers aimed to improve the accuracy and speed of diagnosis, especially in complex cases where isolated symptoms might be misleading.

The Impact:

The context-aware AI diagnostic tool demonstrated a significantly higher accuracy rate in identifying the rare disease compared to traditional methods and earlier AI models that lacked comprehensive contextual input. It was also able to flag potential risks and complications that might have been overlooked otherwise. This case study underscores the critical role of context engineering in high-stakes applications like medical diagnosis, where a holistic understanding of the patient’s situation can lead to more timely and effective treatments.

Key Insight: Context engineering, by enabling a holistic view of a patient’s health and history, is crucial for improving the accuracy and reliability of AI in critical fields like medical diagnosis.

The Future of AI is Contextual

The future of AI is not about building bigger models; it’s about building smarter ones. And a smarter AI is one that can understand and leverage the richness of context, just as humans do. From a human-centered perspective, context engineering is the practice that makes AI more useful, more reliable, and more deeply integrated into our lives in a way that truly helps us. By moving beyond simple prompts and isolated data points, we can create AI systems that are not just powerful tools, but truly intelligent and invaluable partners. The work of bridging the gap between isolated data and meaningful context is where the next great wave of AI innovation will emerge, and it is a task that will demand our full attention.

Image credit: Pexels

Content Authenticity Statement: The topic area and the key elements to focus on were decisions made by Braden Kelley, with help from Google Gemini to shape the article and create the illustrative case studies.

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

Building a Business with Feelings and Emotions

Leave a reply

Building a Business with Feelings and Emotions

GUEST POST from Mike Shipulski

If you use your sane-and-rational lenses and the situation doesn’t make sense, that’s because the situation is not governed by sanity and rationality. Yet, even though there’s a mismatch between the system’s behavior and sane-and-rational, we still try to understand the system through the cloudy lenses of sanity and rationality.

Computer programs are sane and rational; Algorithms are sane and rational; Machines are sane and rational. Fixed inputs yield predicted outputs; If this, then that; Repeat the experiment and the results are repeated. In the cold domain of machines, computer programs and algorithms you may not like the output, but you’re not surprised by it.

But businesses are not run by computer programs, algorithms and machines. Businesses are run by people. And that’s why things aren’t always sane and rational in business.

Where computer programs blindly follow logic that’s coded into them, people follow their emotions. Where algorithms don’t decide what to do based on their emotional state, people do. And where machines aren’t afraid to try something new, people are.

When something doesn’t make sense to you, it’s because your assumptions about the underlying principles are wrong. If you see things that violate logic, it’s because logic isn’t the guiding principle. And if logic isn’t the guiding principle, the only other things that could be driving the irrationality are feelings and emotions. But if you think the solution is to make the irrational system behave rationally, be prepared to be perplexed and frustrated.

The underpinnings of management and leadership are thoughts, feelings and emotions. And, thoughts are governed by feelings and emotions. In that way, the currency of management and leadership is feelings and emotions.

If your first inclination is to figure out a situation using logic, don’t. Logic is for computers, and even that’s changing with deep learning. Business is about people. When in doubt, assess the feelings and emotions of the people involved. And once you understand their thoughts and feelings, you’ll know what to do.

Business isn’t about algorithms. Business is about people. And people respond based on their emotional state. If you want to be a good manager, focus on people’s feelings and emotions. And if you want to be a good leader, do the same.

Image credit: Pixabay

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

The Most Powerful Question

Leave a reply

The Most Powerful Question

GUEST POST from Mike Shipulski

Artificial intelligence, 3D printing, robotics, autonomous cars – what do they have in common? In a word – learning.

Creativity, innovation and continuous improvement – what do they have in common? In a word – learning.

And what about lifelong personal development? Yup – learning.

Learning results when a system behaves differently than your mental model. And there four ways make a system behave differently. First, give new inputs to an existing system. Second, exercise an existing system in a new way (for example, slow it down or speed it up.) Third, modify elements of the existing system. And fourth, create a new system. Simply put, if you want a system to behave differently, you’ve got to change something. But if you want to learn, the system must respond differently than you predict.

If a new system performs exactly like you expect, it isn’t a new system. You’re not trying hard enough.

When your prediction is different than how the system actually behaves, that is called error. Your mental model was wrong and now, based on the new test results, it’s less wrong. From a learning perspective, that’s progress. But when companies want predictable results delivered on a predictable timeline, error is the last thing they want. Think about how crazy that is. A company wants predictable progress but rejects the very thing that generates the learning. Without error there can be no learning.

If you don’t predict the results before you run the test, there can be no learning.

It’s exciting to create a new system and put it through its paces. But it’s not real progress – it’s just activity. The valuable part, the progress part, comes only when you have the discipline to write down what you think will happen before you run the test. It’s not glamorous, but without prediction there can be no error.

If there is no trial, there can be no error. And without error, there can be no learning.

Let’s face it, companies don’t make it easy for people to try new things. People don’t try new things because they are afraid to be judged negatively if it “doesn’t work.” But what does it mean when something doesn’t work? It means the response of the new system is different than predicted. And you know what that’s called, right? It’s called learning.

When people are afraid to try new things, they are afraid to learn.

We have a language problem that we must all work to change. When you hear, “That didn’t work.”, say “Wow, that’s great learning.” When teams are told projects must be “on time, on spec and on budget”, ask the question, “Doesn’t that mean we don’t want them to learn?”

But, the whole dynamic can change with this one simple question – “What did you learn?” At every meeting, ask “What did you learn?” At every design review, ask “What did you learn?” At every lunch, ask “What did you learn?” Any time you interact with someone you care about, find a way to ask, “What did you learn?”

And by asking this simple question, the learning will take care of itself.

Image credit: Pixabay

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

Innovation is Unknowable Not Uncertain

Leave a reply

Innovation is Unknowable Not Uncertain

GUEST POST from Mike Shipulski

Where’s the Marketing Brief? In product development, the Marketing team creates a document that defines who will buy the new product (the customer), what needs are satisfied by the new product and how the customer will use the new product. And Marketing team also uses their crystal ball to estimate the number of units the customers will buy, when they’ll buy it and how much they’ll pay. In theory, the Marketing Brief is finalized before the engineers start their work.

With innovation, there can be no Marketing Brief because there are no customers, no product and no technology to underpin it. And the needs the innovation will satisfy are unknowable because customers have not asked for the them, nor can the customer understand the innovation if you showed it to them. And how the customers will use the? That’s unknowable because, again, there are no customers and no customer needs. And how many will you sell and the sales price? Again, unknowable.

Where’s the Specification? In product development, the Marketing Brief is translated into a Specification that defines what the product must do and how much it will cost. To define what the product must do, the Specification defines a set of test protocols and their measurable results. And the minimum performance is defined as a percentage improvement over the test results of the existing product.

With innovation, there can be no Specification because there are no customers, no product, no technology and no business model. In that way, there can be no known test protocols and the minimum performance criteria are unknowable.

Where’s the Schedule? In product development, the tasks are defined, their sequence is defined and their completion dates are defined. Because the work has been done before, the schedule is a lot like the last one. Everyone knows the drill because they’ve done it before.

With innovation, there can be no schedule. The first task can be defined, but the second cannot because the second depends on the outcome of the first. If the first experiment is successful, the second step builds on the first. But if the first experiment is unsuccessful, the second must start from scratch. And if the customer likes the first prototype, the next step is clear. But if they don’t, it’s back to the drawing board. And the experiments feed the customer learning and the customer learning shapes the experiments.

Innovation is different than product development. And success in product development may work against you in innovation. If you’re doing innovation and you find yourself trying to lock things down, you may be misapplying your product development expertise. If you’re doing innovation and you find yourself trying to write a specification, you may be misapplying your product development expertise. And if you are doing innovation and find yourself trying to nail down a completion date, you are definitely misapplying your product development expertise.

With innovation, people say the work is uncertain, but to me that’s not the right word. To me, the work is unknowable. The customer is unknowable because the work hasn’t been done before. The specification is unknowable because there is nothing for comparison. And the schedule in unknowable because, again, the work hasn’t been done before.

To set expectations appropriately, say the innovation work is unknowable. You’ll likely get into a heated discuss with those who want demand a Marketing Brief, Specification and Schedule, but you’ll make the point that with innovation, the rules of product development don’t apply.

Image credit: Unsplash

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

Innovation or Not – SpinLaunch

1 Reply

Innovation or Not - SpinLaunch

GUEST POST from Art Inteligencia

In the fast-paced world of space exploration, innovation is a driving force that propels new companies and ideas into the spotlight. One such company is SpinLaunch, which is making waves with its novel approach to launching payloads into space. But what sets SpinLaunch apart, and how do we assess whether its approach is truly an innovation or not?

The Concept Behind SpinLaunch

SpinLaunch is taking a radically different approach to space launch by using a kinetic energy-based system rather than traditional rocketry. Their technique involves a high-speed rotating arm that builds up momentum and catapults a payload to the edge of space, drastically reducing the need for fuel and cutting down on costs. This approach is not only cost-effective but also environmentally friendly, addressing two significant pain points in the space industry.

Key Criteria for Innovation Assessment

Novelty: Is the concept fresh and previously unexplored?
Feasibility: Can the technology be realistically executed?
Impact: What benefits does the innovation provide to the industry and society?
Scalability: Can the idea grow and adapt to broader applications?

Case Study: Assessing SpinLaunch

Novelty

SpinLaunch undoubtedly introduces a novel approach to space launches. Traditional methods rely heavily on chemical propulsion. In contrast, SpinLaunch’s kinetic system stands out by leveraging physics in a way that hasn’t been commercially applied to space launches before.

Feasibility

The technical feasibility of SpinLaunch’s idea has been demonstrated through successful suborbital launches, proving that their kinetic system can indeed hurl payloads into space. However, the transition from suborbital to orbital flights will be the true test of feasibility. Critical engineering challenges remain, particularly related to the G-forces sustained by payloads during launch.

Impact

SpinLaunch has the potential to revolutionize the space industry by making launches significantly cheaper and more frequent. The environmental benefits of reducing fuel consumption cannot be understated either. If successfully scaled, the impact would reach beyond cost — it could democratize access to space.

Scalability

Currently, SpinLaunch is focused on small to medium-sized payloads. For scalability, the company must expand its capabilities to accommodate larger satellites and potentially human passengers. Adapting the technology for broader applications will be essential.

Conclusion: Is SpinLaunch an Innovation?

SpinLaunch exhibits the hallmarks of a true innovation. By addressing cost, environmental impact, and frequency of launches, it provides substantial benefits to the space industry. However, the road to demonstrating full potential is fraught with engineering and market challenges. Yet, the novelty and promise of their approach cannot be ignored.

Here is a 40 minute documentary that dives deep into the engineering, problem solving and innovation approach:

Opportunities for Expansion

To strengthen the case for SpinLaunch as an innovation, future assessments could involve the impact on related industries such as satellite manufacturing. More real-world data from further launches will offer insights into long-term feasibility and environmental impact. Engaging with regulators and potential partners early will be crucial to addressing scalability challenges.

Revision & Expansion

The ongoing journey of SpinLaunch should be closely monitored. As the company progresses, it should aim to address:

Risk Management: How can the company mitigate potential risks associated with high G-force impacts on sensitive equipment?
Regulatory Hurdles: Navigating international laws and space treaties will be essential as SpinLaunch aims for global reach.
Commercial Partnerships: Collaborations with established aerospace companies could fast-track development and market entry.

The future of SpinLaunch lies in its ability to resolve these emerging challenges while maintaining its innovative edge, positioning the company as a potential leader in transforming space access.

So, what do you think? Innovation or not?

Image credit: SpinLaunch

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