Ordered to Deploy AI, Excluded From the Room: The Real CX Leadership Challenge

Nine in 10 service leaders have been told to deploy AI. Far fewer were asked how. The gap between the instruction and the authority is where customer experience leadership is being decided this year.

Something changed in the way executive teams talk about AI this year, and most customer experience leaders felt it before they could name it. The argument about whether to deploy stopped. The budget appeared. The deadline appeared. And somewhere in that shift, the question of who decides how it lands on customers quietly went unasked.

I keep meeting service and experience leaders in the same position. They have the mandate. They have the visibility. They will carry the blame when a customer is trapped in a loop at 11 at night with no way through to a person. What they do not have is authority over the systems, the budget, the job design or the definition of success that will produce that moment.

The industry keeps misdiagnosing this. It gets called a skills gap, so we send people on courses. It gets called change resistance, so we run engagement sessions. It gets called an AI literacy problem, so we buy more training. None of that touches the actual constraint, which is that the people accountable for customer outcomes were not in the room when the decisions that determine those outcomes were made.

That room is being furnished right now. The seating plan for who owns AI-era work is being drawn in 2026, and it will hold for years. This piece lays out what the evidence says about who currently holds the pen,

Why Service Leaders Own AI Outcomes but Not the Decisions Behind Them

Ninety-one percent of customer service and support leaders report pressure from their own executives to implement AI. That figure comes from a survey of 321 service leaders run last October by Gartner.

Be precise about what that number measures. It is not adoption. It is instruction. Nine in 10 people running service operations were handed a decision made somewhere else.

Look at what they now plan to do with it. Nearly 80% will move agents into new roles. Eighty-four percent will rewrite hiring profiles. Fifty-eight percent will turn agents into knowledge managers, because self-service collapses without maintained content behind it.

Every one of those is a workforce decision. Almost none of them belongs to the person who will be held responsible for the result. Job architecture sits with HR. Hiring profiles sit with HR. The knowledge platform sits with IT. The budget sits with the CIO or a central AI function.

So the service leader owns the outcome and controls none of the machinery that produces it. That is not leadership. That is exposure with a job title attached.

Look also at the stated priorities for the year: better customer satisfaction, better efficiency, better self-service containment. Anyone who has run a service floor knows those three pull against each other, and they pull hardest at the exact moment a customer has a problem the machine cannot solve. You do not settle that on a priority slide. You settle it by deciding when the machine must hand over to a person. That decision usually gets made three functions away from whoever will answer for it.

Why Does Executive Pressure to Deploy AI Leave Service Leaders Exposed?

Service leaders are held accountable for AI outcomes, but the workforce decisions that follow sit with HR, IT and the CIO: role changes, hiring profiles, knowledge platforms and budget. They own the result without controlling the machinery that produces it.

What the AI Governance Proof Gap Means for Customer Accountability

Here is the part that should worry every executive team, and it comes from a different place entirely. Grant Thornton surveyed 950 senior leaders across 10 industries between late February and mid-March this year. Seventy-eight percent lacked strong confidence that they could pass an independent AI governance audit within 90 days.

Read that again. Three-quarters of boards have signed off major AI investment. Forty-eight percent never set governance expectations. Forty-six percent never made AI risk a standing item. Money approved, accountability never assigned.

The same study puts a number on the split I see in every client conversation. Chief information and technology officers are five times more likely than chief operating officers to say the workforce is ready for AI. On agentic risk the gap inverts: 54% of operating leaders worry about regulatory and compliance exposure, against 20% of their technology counterparts.

Those are not two opinions. Those are two different organizations, described by people sitting at the same table.

And the exposure is already live. Nearly three in four organizations have given agentic AI access to their data and processes. One in five has a tested incident response plan for when it fails. The researchers call the result a proof gap, and their own summary is blunter than anything I would write: leadership deployed AI without defining who owns the outcomes.

That sentence is the whole problem. Not skills. Not literacy. Not vision. Ownership.

Related Article: Agentic Customer Experience: The CX Architecture Built for the World Customers Actually Live In

What Does the AI Governance Proof Gap Reveal About Outcome Ownership?

Most senior leaders doubt they could pass an independent AI governance audit within 90 days, and only one in five has a tested incident response plan. Deployment has moved faster than anyone's assigned accountability for outcomes.

Why the CHRO-CIO Model for AI Work Redesign Leaves Out the Customer

Three days before I wrote this, Gartner published its answer to who owns the redesign of work for the AI era. The answer is the CHRO and the CIO, jointly, across five capabilities: AI ethics and governance, skills-based talent management, total cost of work, human-agent job redesign, and human-AI performance management. By 2029, the forecast says, 30% of organizations will run blended HR-IT teams to do it.

Every one of those five decides something a customer will feel. Job redesign decides who handles a complaint. Performance management decides what an agent is rewarded for doing. Total cost of work decides how many humans are still reachable at seven on a Friday evening.

The customer-facing function is not named as a party to any of it.

I do not think the analysis is wrong. Work redesign genuinely sits between people and technology. The warning attached to it is the honest part: role confusion, decisions taken in silos without business alignment and one-sided transformation where HR optimises for people while IT optimises for innovation and neither is accountable for the result.

Add the missing variable and the risk sharpens considerably. When two functions measured on workforce cost and technology delivery redesign the work, the customer outcome becomes whatever is left over. Nothing in that structure makes anyone answerable for what the customer actually receives. Residuals do not get defended in budget meetings.

Ownership fragmentation is not a niche complaint either. In a study of more than 350 senior leaders at companies above 1,000 employees published by the measurement platform Larridin, 58% named unclear or fragmented ownership as the single biggest barrier to measuring AI performance at all. Treat vendor-commissioned research with the caution it deserves, then notice it points the same direction as everything above.

Why Does Joint HR-IT Ownership of AI Work Redesign Put Customer Outcomes at Risk?

Job redesign, performance management and total cost of work all shape what customers experience, yet the customer-facing function isn't a named party. When HR and IT optimize for workforce cost and delivery, customer outcomes become a residual nobody defends.

Where AI Ownership Breaks Down Between Service, HR and IT

Editor's note: The following table highlights the most important lessons, actions and strategic considerations emerging from the split between AI accountability and AI authority in customer service.

Key Area	What Happened	Why It Matters	Recommended Action
Deployment Mandate	Nine in 10 service leaders face executive pressure to implement AI	The mandate arrived without authority over roles, platforms or budget	Tie AI accountability to explicit configuration rights
Governance Readiness	Most leaders doubt they could pass an AI governance audit in 90 days	Investment was approved before outcome ownership was assigned	Require named owners and tested incident response before launch
Work Redesign Ownership	Gartner assigned AI work redesign jointly to the CHRO and CIO	Customer outcomes become a residual no one defends	Secure a formal CX seat in HR-IT redesign governance
Frontline Exposure	Agentic AI lands first in support, where governance is least mature	Failures surface as complaints, churn and regulatory attention	Claim the escalation path and disclosure standard in writing
Success Metrics	Containment and deflection stand in for resolution	Whoever defines success controls program economics	Define what counts as a resolved contact and reject deflection as a primary KPI

Why Customer Support Absorbs Agentic AI Failures First

Service is where the consequences arrive earliest, and there is a reason for that beyond bad luck.

Deloitte's enterprise AI research identifies customer support as the function where agentic AI is expected to deliver its highest impact. The same work finds that only one company in five has a mature governance model for autonomous agents, and that 37% of organisations are still applying AI at surface level with little or no change to the underlying process.

Put those together. The technology goes first into the function with the most customer contact, the least governance maturity and frequently no process redesign underneath it. Then the invoice for that decision is presented to the service organization in the form of complaints, churn and regulatory attention.

This is also where the ambiguity gets expensive rather than merely annoying. When a bot invents a policy or traps someone in a loop, the escalation arrives at service. The escalation threshold, the disclosure wording and the human fallback were configured elsewhere. Analysts at Forrester describe the same pattern from the other end: plenty of organizations can point to AI initiatives, far fewer can say why those initiatives exist, what success looks like, or where human oversight still belongs.

That is not a technology gap. It is an unowned decision, and it becomes visible only when a customer is standing in it.

Why Do Agentic AI Failures Surface in Customer Support First?

Customer support is the expected highest-impact function for agentic AI, but few companies have mature agent governance. Escalation thresholds, disclosure wording and human fallbacks are configured elsewhere, while the complaints land in service.

The Decision Rights CX Leaders Should Claim Before the Next AI Release

Stop asking for influence. Influence is what you are given instead of authority. Ask for decision rights, in writing, before the next release.

• Claim the escalation path. When the machine must hand a customer to a person, and how quickly. This is the single highest-leverage control in any deployment.
• Claim the disclosure standard. What the customer is told about what they are dealing with, in what words, at what moment. In Europe this is now a legal duty as well as a design choice.
• Claim the definition of a resolved contact. Not closed. Not deflected. Resolved. Whoever writes that definition controls the economics of the whole programme, and if you do not write it, finance will.
• Claim a release veto tied to named quality conditions rather than to opinion, so it survives the first argument with a delivery date.

Then refuse three things, and refuse them early. Refuse deflection as a primary performance indicator. Refuse accountability for outcomes on systems you have no rights to configure — either the rights travel with the accountability or the accountability moves. And refuse any pilot without a written stop condition, because a pilot that cannot fail is not a pilot. It is a rollout with better manners.

One prior question sits underneath all of it: whether AI is the right instrument here at all. A governance problem does not become an AI problem because an executive used the word in a board meeting.

Which AI Decision Rights Should Customer Experience Leaders Put in Writing?

CX leaders should claim four rights: the escalation path, the disclosure standard, the definition of a resolved contact, and a release veto tied to named quality conditions. They should also refuse three things: deflection as a primary KPI, accountability without configuration rights, and pilots without stop conditions.

How CX Leaders Earn a Seat in AI Work Redesign

Nobody is going to hand this over. The three-year window is already visible in the forecasts, and when it closes the operating model of customer-facing work will have been rebuilt by whoever showed up. Functions absent from that room will spend the next decade administering a design they had no part in.

The way in is not fluency in AI vocabulary. Half the executive committee already has that and it has bought them nothing. The way in is being the only person in the building who can answer two questions with evidence: what will this cost us in customers, and how will we know before it is irreversible.

Those questions are harder than they look, which is precisely why they are worth owning. Answering the first means connecting a design decision — an escalation threshold, a containment target, a staffing floor — to retention and customer lifetime value rather than to handle time. Answering the second means instrumenting the failure modes before launch, not reconstructing them from complaints afterwards. Do that work and the invitation stops being something you request. It becomes something the organisation cannot function without.

No HR leader can answer those two questions. No CIO can answer them. If the customer leader cannot answer them either, then the seat was never really theirs to lose.

Ninety-one percent were told to deploy AI. The ones who still matter in 2029 will be the minority who insisted on defining what deploying it meant.

Learn how you can join our contributor community.