Editorial
White Paper
The New Publishing Standard in the AI Era and What It Replaces
The Gist
- Not a Tech Failure. AI agent failures — wrong claims approvals, unhonored prices, trapped vulnerable customers — trace back to organizational gaps, not the model itself.
- CX Shouldn't Own It. AI governance is genuinely cross-functional; handing it solely to CX would repeat the same siloing mistake as any other single-function ownership.
- CX's Real Job Is the Question. CX's unique value is carrying "how does this affect the customer?" into a room where data, legal, and ops teams are each optimizing for their own metric.
- Start Before the Invitation. CX leaders should build the CTO/CAIO relationship, bring evidence of customer-impact patterns, and ask the customer question early — before governance formally convenes.
An AI agent tells a customer their claim is approved. It wasn't. Or it quotes a price the company can't honor. Or it traps a vulnerable customer — elderly, disabled, in distress — in an automated loop they can't get out of, with no human to turn to. By the time anyone notices, the damage is done — to the customer, and to the trust the business spent years building.
When you trace these failures back, the cause is rarely the model itself. It's something upstream: data that was biased or incomplete, a team that didn't have the AI literacy to catch the problem, a key stakeholder who was never in the decision, no guardrails, unclear objectives, or no one checking whether the model still worked six months after it launched.
None of those are technology problems. They're organizational ones. And that changes who needs to be involved in fixing them.
The Research Is Clear: This Isn't a Tech Problem
The data on AI failure is remarkably consistent on one point — the algorithm is rarely the issue.
Gartner has projected that through 2026, organizations “will abandon 60% of AI projects unsupported by AI-ready data”. McKinsey's research points the same direction: the organizations getting real returns from AI aren't the ones with the best models — they're the ones that redesigned their data and workflows first.
And the customer is increasingly on the receiving end when this goes wrong. In the same report, McKinsey found that “51% of organizations using AI reported at least one negative AI incident”. Almost one-third of respondents reported issues from inaccurate outputs.
So the failures are organizational, and their consequences land on the customer. Which raises a question CX leaders should be asking: if this is an organizational problem with customer consequences, where is CX in the process that's supposed to prevent it?
Too often, nowhere.
Related Article: The Scale Model: A CX Playbook for AI That Doesn't Backfire
What Percentage of AI Projects Fail Due to Data Readiness?
Gartner has projected that through 2026, organizations will abandon 60% of AI projects unsupported by AI-ready data — evidence that failure is upstream of the model.
Why Speed-to-Ship Culture Undermines AI Governance
The gap exists partly because everything about AI pushes toward speed.
Chad Sanderson, CEO of Gable.ai, in Secoda’s State of Data Governance in 2025 report, named the underlying tension: "The incentive of one group is to slow down and be careful and do the right thing, and the incentive of the other group is to ship as quickly as possible."
Right now, ship-fast usually wins. And when it does, the careful questions — including the one about the customer — get skipped in the rush to launch.
This tension is perfectly illustrated in this Marketoonist cartoon:
Misalignment is creating this issue. While misalignment isn’t anything new, it’s exposed thanks to AI’s speed at scale; customers notice and trust worsens.
Why Does Speed-to-Ship Culture Undermine AI Governance?
Gable.ai CEO Chad Sanderson describes a structural tension: one group's incentive is to move carefully, the other's is to ship fast — and ship-fast usually wins, skipping the customer-impact question.
CX Shouldn't Own AI Governance
Here's where I'll be clear about something that runs against a lot of current commentary: I don't think AI governance is CX's job to own.
Governing AI is a genuinely cross-functional responsibility — data, legal, security, risk, product, operations. Handing it to CX would be the same mistake as handing it to any single function: it would silo a problem that is, by its nature, shared. CX doesn't have the mandate, and shouldn't claim it.
But not owning something isn't the same as not belonging in it.
Who Should Own AI Governance Instead of CX?
No single function should own it — data, legal, security, risk, product and operations share responsibility, since AI failures emerge from the seams between functions, not within one of them.
CX's Real Role: The Uniter — and the Customer's Voice
CX has one quality almost no other function has: it touches every part of the business. Product, operations, marketing, pricing, service — the customer experience runs through all of them. CX knows how alignment (or misalignment) impacts customers. That cross-functional reach is exactly what AI governance needs, because the failures come from the seams between functions, not from within any one of them.
So the CX leader is well-positioned to be a uniter — helping convene the cross-functional group that governs AI, whether you call it an AI Center of Excellence or something else. And to be clear, that's not a new bureaucracy or a platform to buy. It's a room: the right people — data, risk, legal, product, operations and CX — aligned on how AI gets deployed, with someone accountable for the questions that fall between them.
And CX's specific job in that room is to carry the question that keeps getting skipped: how does this affect the customer?
That question sounds obvious. It almost never gets asked in the moment a model ships. The data team is focused on accuracy. Legal is focused on compliance. Operations is focused on efficiency. Each is doing its job. But no one in the room is accountable for the customer's experience of the system — unless CX is there to be.
When an AI agent produces a wrong answer, or an adverse outcome, or quietly erodes customer trust, it's almost always because that question wasn't owned by anyone at the point the decision was made. CX in the room is how it gets owned.
What Question Should CX Carry Into AI Governance Discussions?
CX's job is to keep asking "how does this affect the customer?" — the question each function skips while focused on its own metric (accuracy, compliance, efficiency).
What CX Brings That No One Else Does
This isn't about CX policing the other functions. It's about completing the picture.
The data team can tell you whether the model is accurate. CX can tell you whether "accurate" is the same as "good for the customer" — because sometimes it isn't. Legal can tell you whether a disclosure is compliant. CX can tell you whether a customer will understand it. Operations can tell you whether removing a human touchpoint cuts cost. CX can tell you what it does to trust.
That translation — from functional metric to customer consequence — is what CX uniquely provides. And it's precisely what's missing when AI fails a customer.
How Does CX's Contribution Differ From Legal or Operations?
Legal confirms a disclosure is compliant; CX confirms a customer will understand it. Operations confirms cutting a human touchpoint saves cost; CX confirms what it does to trust.
3 Moves CX Leaders Can Make Before AI Governance Convenes
Editor's note: The following table highlights the most important lessons, actions and strategic considerations emerging from what CX leaders can do now to shape AI governance.
| Action | What It Means | Why It Matters |
|---|---|---|
| Build the relationship with whoever owns AI | Get to know your CTO or CAIO now to understand how they think about AI risk | Influence is built before the decision, not requested after it |
| Bring evidence, not opinions | Show up with the customer-impact patterns you're already seeing — where AI helps, where it erodes trust | The functions running governance respond to evidence, and CX has it |
| Ask the customer question early and often | In every AI conversation, ask "how does this affect the customer?" until the room expects it | You don't need the mandate to be the customer's voice — you need to be useful enough that the room can't happen without you |
What Should CX Leaders Do Before a Governance Room Exists?
Build the relationship with whoever owns AI (CTO or CAIO), bring evidence of customer-impact patterns instead of opinions, and ask the customer question in every AI conversation until it's expected.
The AI Governance Imperative for CX Leaders
So the AI governance imperative for CX leaders isn't to own the function. It's to make sure CX is in the room where AI decisions get made — as the uniter who helps bring the functions together, and as the voice accountable for the customer inside a process that will otherwise optimize everything except them.
AI governance will happen with or without CX. The question is whether the customer has a seat at the table when it does — or whether they only show up later, in the incident report.
Learn how you can join our contributor community.
