by Braden Kelley and Art Inteligencia
What Are the Trust Pillars for Agentic CX? (Short Answer)
Six trust pillars for agentic customer experience: (1) Clarity, (2) Competence, (3) Control, (4) Care, (5) Consent, and (6) Accountability. Agents that act — refund, rebook, route, negotiate, commit — change the trust contract. Containment, deflection, and “the model said so” are not design. Soft landings design all six before scale.
Agentic CX earns trust when customers can see the actor, finish the job, stay in control, feel cared for, consent to the scope of action, and still find a human accountable when the agent gets it wrong.
Why Six Pillars — Not Only Automation Metrics?
I keep watching teams ship agentic power as if a smarter FAQ were enough. It isn’t. Agentic CX is delegated action. Customers will forgive imperfect speed. They will not forgive betrayal — opacity, trapped loops, brand-first optimization, no undo, and nobody who can be asked why.
The original four — Clarity, Competence, Control, Care — still hold. I unpacked that contract in When AI Agents Act on Your Behalf. Two pillars complete it when authority is real: Consent (what they authorized the agent to do) and Accountability (who owns recovery when it fails). Autonomy without those six is a hard landing with a better chatbot costume.
For the broader product lens beyond agents, see 11 Principles for Designing Trust (Not Just Usability). For how leaders must own agentic scenarios across the enterprise, see 5 Scenarios for Agentic Organizations.
| Pillar | Customer question | Hard landing tell |
|---|---|---|
| 1. Clarity | Who is acting and what can it do? | “Wait — was that a bot?” after damage |
| 2. Competence | Will this finish without making me restart? | Loops; retelling tax |
| 3. Control | Can I stop, undo, or reach a human? | Trapped; punished for escalating |
| 4. Care | Is this for me or for containment? | Cheap resolution that costs dignity |
| 5. Consent | Did I authorize this scope of action? | Surprises in money, data, commitments |
| 6. Accountability | Who owns this when it fails? | “The system decided”; no recovery power |
Ship the agent’s power only as fast as you can design these six.
1. What Is Clarity in Agentic Customer Experience?
Pillar: Make the actor, scope, and limits visible — human, system, or agent — before and during action.
When agents act: Customers must know they are delegating, not chatting with a FAQ costume.
Hard landing: Hidden automation; surprise commitments; “Was that a bot?” after the charge.
Soft-landing move: Disclose identity, capabilities, and boundaries in the channel; update when scope expands.
Tell: Post-action confusion; chargebacks; “I didn’t know it could do that.”
2. What Is Competence When Agents Act?
Pillar: Resolve completely across steps; carry memory; don’t loop people through the same dead end.
When agents act: Multi-step work is the product — not a clever first reply.
Hard landing: Partial completion; lost context; restart tax; endless clarification.
Soft-landing move: End-to-end completion metrics; context that travels; escalate before the third loop.
Tell: Repeat contacts; “I already told you”; high containment, low resolution. Efficient misery often starts here — see 8 Service Design Mistakes That Create Efficient Misery.
3. Why Does Control Matter More When Agents Have Authority?
Pillar: Real agency includes pause, reverse, override, and human help without punishment.
When agents act: Authority without an exit is coercion with better UX.
Hard landing: No undo on money or data moves; escalation buried; AHT and containment punish the handoff.
Soft-landing move: One-tap stop/undo for high-stakes actions; a human path with teeth; a dual scorecard so green containment cannot close a red trust review.
Tell: Customers force a channel-switch to escape the agent; human agents inherit angry cleanup.
4. What Does Care Mean for Agentic CX?
Pillar: At the decision point, design for the customer’s job and dignity — not only conversion, deflection, or cost.
When agents act: The agent’s objective function is the brand promise in motion.
Hard landing: Containment KPIs; upsell in distress; “resolved” tickets that leave the job unfinished.
Soft-landing move: Objective functions and policies that prefer make-right over cheap close; care metrics beside automation rate.
Tell: High script scores, low loyalty; viral “the bot wouldn’t help” stories. Soft landings split human and machine work on purpose — see The AI Soft Landing.
5. Why Is Consent a Trust Pillar for Agentic CX?
Pillar: Explicit, reversible permission for what the agent may do with money, data, identity, and commitments — matched to stakes.
When agents act: Delegation is a trust decision, not a default toggle buried in settings.
Hard landing: Silent scope creep; agents that “helpfully” act beyond what was asked; creepy personalization treated as permission.
Soft-landing move: Scope prompts for high-stakes bands; easy revoke; minimize data and action to the job.
Tell: “I never authorized that”; regret after auto-actions; consent theater in the fine print.
6. What Is Accountability When the Agent Gets It Wrong?
Pillar: A named human (or role) and a powered recovery path when the agent errs — explainability plus make-right, not a shrug.
When agents act: “The system decided” is not a brand. Someone must still be askable.
Hard landing: Unowned model choices; recovery as exception hell; liability parked on the customer.
Soft-landing move: Named accountable owner per journey; recovery journeys equal to the primary flow; explain why in plain language; log decisions for appeal.
Tell: No one to escalate to; make-right requires a second odyssey; trust dies at the seam.
What Should You Ask Before Scaling Agentic CX?
Six questions before the next agentic CX release:
- Is the actor and scope clear?
- Can the median journey finish without a loop trap?
- Can the customer stop, undo, and reach a human without punishment?
- Does the objective function optimize for their interest?
- Was consent earned for this band of action?
- Who owns recovery — with power — when it fails?
Mantra: Don’t scale the agent’s authority faster than you can design Clarity, Competence, Control, Care, Consent, and Accountability.
FAQ: Trust Pillars for Agentic Customer Experience
What are the trust pillars for agentic CX?
The six trust pillars for agentic CX are Clarity, Competence, Control, Care, Consent, and Accountability — the contract customers need when AI agents act on their behalf, not only chat.
How do you design trust for AI agents?
Design trust for AI agents by disclosing who is acting and what they can do, finishing jobs with context, giving stop/undo/human paths, optimizing for the customer’s interest, earning consent for the scope of action, and naming who owns powered recovery when the agent fails.
What is consent in agentic customer experience?
Consent in agentic customer experience is explicit, reversible permission for what the agent may do with money, data, identity, and commitments — matched to stakes — not a buried default that expands the agent’s authority without the customer knowing.
Why isn’t automation enough for CX trust?
Automation is not enough for CX trust because agentic systems take actions that can help or harm. Speed and containment without clarity, control, care, consent, and accountability feel like betrayal — and customers forgive friction more readily than betrayal.
How do you measure agentic CX trust?
Measure agentic CX trust with completion without loops, undo and human-reach success, consent revoke rates, recovery time and power after agent error, and whether customers understood who acted — not automation or containment rate alone.
Image credits: Pixabay
Content Authenticity Statement: The topic area, key elements to focus on, etc. were decisions made by Braden Kelley, with a little help from Google Gemini and Cursor to clean up the article, add images and create infographics.
Sign up here to get Human-Centered Change & Innovation Weekly delivered to your inbox every week.