Nine in Ten CX Leaders Ordered to Deploy AI Without Authority Over Systems or Budget, Research Finds

A cluster of industry research published in 2025 and 2026 reveals that most customer experience (CX) and customer service leaders have been directed by their executives to implement artificial intelligence, yet control over the workforce decisions, platforms and budgets that shape those deployments sits elsewhere — chiefly with HR, IT and the CIO.

Mandate Without Authority

A Gartner survey of 321 customer service and support leaders, conducted in October 2025, found that 91% reported executive pressure to implement AI. According to an analysis published by CMSWire on 29 September 2026, the survey also showed that nearly 80% of those leaders plan to move agents into new roles, 84% will rewrite hiring profiles, and 58% will turn agents into knowledge managers. The analysis notes that job architecture and hiring profiles sit with HR, while knowledge platforms and budgets sit with IT or a central AI function — leaving the service leader accountable for outcomes without control over the machinery that produces them.

Governance Readiness Gaps

A separate Grant Thornton survey of 950 senior leaders across ten industries, conducted between late February and mid-March 2026, found that 78% lacked strong confidence they could pass an independent AI governance audit within 90 days. The same study found that 48% of boards had never set governance expectations for AI and 46% had never made AI risk a standing agenda item. On agentic risk specifically, the Grant Thornton data showed that 54% of operating leaders expressed concern about regulatory and compliance exposure, compared with 20% of their technology counterparts. The CMSWire analysis also cited the finding that nearly three in four organisations have given agentic AI access to their data and processes, while only one in five has a tested incident response plan.

Work Redesign Ownership

A Gartner report published in August 2025 assigned joint ownership of AI-era work redesign to the CHRO and CIO across five capabilities: AI ethics and governance, skills-based talent management, total cost of work, human-agent job redesign, and human-AI performance management. Gartner forecast that by 2029, 30% of organisations will run blended HR-IT teams for this purpose. The CMSWire analysis noted that no customer-facing function is named as a party to any of these five areas, meaning customer outcomes become, in its framing, "whatever is left over."

A study of more than 350 senior leaders at companies with more than 1,000 employees, published by measurement platform Larridin, found that 58% named unclear or fragmented ownership as the single biggest barrier to measuring AI performance. The CMSWire analysis flagged that the Larridin research is vendor-commissioned and should be treated with corresponding caution, while observing that its direction aligns with the other evidence cited.

Customer Support as the Primary Failure Point

Deloitte enterprise AI research, also cited in the CMSWire analysis, identifies customer support as the function where agentic AI is expected to deliver its highest impact. The same Deloitte work found 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. Analysts at Forrester, also referenced in the piece, are described as finding that many organisations can point to AI initiatives but far fewer can explain why those initiatives exist, what success looks like, or where human oversight still belongs.

Recommended Response

The CMSWire analysis argues that CX leaders should seek explicit decision rights in writing before any deployment, specifically over the escalation path, disclosure standard, definition of a resolved contact, and a release veto tied to named quality conditions. It also advises leaders to refuse deflection as a primary performance indicator, to reject accountability for outcomes on systems they cannot configure, and to require a written stop condition for any pilot programme.