From Assistive to Agentic: What the Largest AI in CX Study Reveals for Leaders in 2026

CX Network publishes one of the most comprehensive annual studies on the use of artificial intelligence in customer experience. The 2026 edition, The State of AI in CX: From Assistive to Agentic, surveyed 342 CX professionals, service leaders, experience designers, analysts, and consultants around the world. The result is a precise picture of where organizations stand, where they are headed, and — most importantly — what separates those building real competitive advantage from those merely keeping pace.

This post summarizes the study's key findings, cross-references them with other relevant sources, and unpacks what these numbers mean for decision-makers.

AI in CX Has Moved from Trend to Infrastructure

For the second consecutive year, AI technologies for operations topped the ranking of trends most likely to shape CX through 2030 (selected by 24% of respondents). Close behind: agentic AI and AI agents (21%), AI-first customer journeys (17%), and conversational AI (16.5%).

That two-year pattern isn't trivial. When the same trend leads a study of this scope for back-to-back years, the signal is clear: this is no longer about experimentation, but structural adoption.

The McKinsey Global Survey on AI, published in early 2025, reinforces this point: organizations reporting consistent returns from AI are those that treat it as infrastructure (embedded in core processes) rather than as isolated projects. That's the same distinction CX Network captures: AI is migrating from "support tool" to "operational layer."

"The flashy part is the chatbot. The back-end is where the real leverage lives."

Investment Plans Confirm the Direction

Budget data backs up what the trend rankings already signaled. When asked about their CX investment priorities for 2026:

• →29% plan to invest in agentic AI and AI agents
• →22% in automation of CX and service functions
• →15% in chatbots and virtual assistants
• →13% in AI and ML for business operations

Even more telling are the spending growth expectations: 46% of respondents plan to increase their Generative AI investment, 40% in agentic AI, and 39% in other AI tools for CX and service.

One figure worth paying close attention to: 68% of organizations are acquiring new AI capabilities through third-party vendors, not internal development. This has direct implications for anyone defining supplier relationships and evaluating technology partners, since it shapes how the company will evolve in AI. Not every organization has the capacity (or budget) to build all AI capabilities in-house, and in many cases it doesn't make strategic sense. But it is worth each company mapping how many vendors can serve it, which areas it wants to prioritize for internal AI capability development, and which ones are "supporting" or non-priority (and therefore appropriate to source from third parties). The technology purchasing decision has become, in large part, a CX decision.

The study also maps who decides: 37% of respondents are part of the decision-making team on AI investments, 20% influence those decisions, and 15% hold exclusive responsibility. The CX function is no longer a passive consumer of technology — it now has a seat at the table where budgets are set.

Your Customers Are Also Using AI to Buy (and That Changes Everything)

One of the study's most strategic findings concerns not the company's AI, but the customer's. A growing share of consumers are now using AI assistants to research products, compare vendors, resolve issues, and make purchasing decisions. Gartner, which coined the concept of machine customers in 2018, projects that AI buying agents will influence up to 20% of purchasing decisions by 2028, and what was once a projection is starting to show up as reality in field research.

The impact on organizations is direct: when an AI assistant evaluates vendors on a customer's behalf, it is not swayed by brand campaigns, relationship history, or commercial narratives, since it reads objective signals (delivery reliability, ease of returns, information consistency, complaint responses, structured presence in the sources that language systems index).

"Discovery is now about how well your business holds up when an AI is evaluating it on behalf of a customer."

This creates a new category of competitive requirement: Generative Engine Optimization (GEO) and Answer Engine Optimization (AEO). Organizations that don't structure their information so that language systems can read, interpret, and accurately recommend them simply drop out of the equation before a human ever sees their name. The content, data, and online credibility layer has become sales infrastructure, not just a marketing exercise.

AI as an Operational Layer: The Cases That Prove It

The study features two case studies that illustrate, with concrete data, what happens when AI in CX is implemented with a clear problem definition and proper governance.

CVS Health

The company created 100,000 AI digital twins (behavioral models built from 2.9 million real responses from over 400,000 participants across more than 200 scenarios). The goal: scale the Voice of the Customer program beyond what traditional surveys could support. The result: simulations that compress weeks of research into hours, with 85–95% accuracy and the ability to test new hypotheses around experience, operations, and service without involving real customers in every cycle.

"We used to have an empty chair in every meeting to represent the customer. We no longer need that."

LoadUp

The on-demand junk removal company faced a structural bottleneck: each quote required significant human review, limiting scale and slowing conversions. By implementing the Kustomer platform, an AI agent now handles 100% of SMS sales inquiries, resolves 33% of those inquiries end-to-end (converting them into scheduled jobs), and has increased customer satisfaction by 5 percentage points. The strategic differentiator, according to Chad Danklef, VP of Operations, wasn't just efficiency — it was the upfront, transparent pricing promise that automation made viable at scale.

These two cases illustrate what other sources, such as the Salesforce State of Service report, also find: the biggest gains from AI in CX come from internal applications (routing, quality analysis, knowledge management, triage) rather than from customer-facing interfaces.

AI Ethics and Governance: From Differentiator to Qualifying Criterion

The study documents a significant shift on the topic of AI governance. In 2025, 48% of organizations reported having no governance models in place, but by 2026 that number dropped to 20%. The progress is notable, yet what changes is not only the volume of policies — it's what's at stake.

The top customer behavior influencing CX planning in 2026, for the first time ever, is awareness of how AI works and uses consumer data (selected by 36% of respondents), surpassing even the demand for speed and convenience that had led this ranking in prior years.

"If an organization's AI ethics aren't machine-readable, they don't exist — at least not for the AI agents increasingly making purchasing decisions on behalf of your customers."

That elevates the governance debate well beyond compliance. ISO/IEC 42001:2023 (the first international standard for AI Management Systems) is beginning to function as a vendor qualification criterion. Organizations that adopt it proactively are positioning themselves to be discovered and preferred by AI buying agents, while those that treat ethics as a communications exercise tend to compete on price alone and get filtered out at machine-processing speed.

The study points to three questions every AI governance policy needs to be able to answer today:

1. What happens when an external agent acts as your customer and something goes wrong?
2. Who bears responsibility when your customer's AI agent purchases incorrectly or exploits a gap in your pricing logic?
3. How do you revoke agent access cleanly, without abandoned credentials creating security vulnerabilities?

These are not theoretical questions — they are operational questions that need answers before the problem surfaces.

What These Findings Mean for Decision-Makers

The study leaves three central implications for leaders:

First: AI in CX requires a defined problem to generate value. The CVS Health and LoadUp cases share one thing: clarity about what was being solved before any technology was chosen. AI deployed without that anchor tends to produce efficiency dashboards that don't translate into better customer experiences.

Second: Customer listening is being redefined. Traditional VoC processes (post-interaction surveys, periodic NPS, focus groups) are being complemented or replaced by continuous analysis, behavioral models, and agentic simulation. Leaders still relying exclusively on analog methods to understand their customers are operating with a structural lag.

Third: The differentiation window for governance is open, but not for long. Today, having a robust and verifiable AI governance policy is a competitive advantage, but as AI buying agents become standard, it will shift to a table-stakes requirement. Those who act now get ahead; those who wait will be playing catch-up.

The IBM Institute for Business Value estimates that 40% of global companies expect to reskill their teams in response to AI over the next three years, and the CX Network study confirms the same shift: 36% of respondents identify reskilling as the single biggest change that agentic AI is bringing to their organizations. The capacity to operate in this environment is not only technological — it's also organizational.

For Those Who Want to Go Deeper

This post covers the study's key highlights, but it's far from exhausting the material. The original report includes detailed analysis on how to embed AI across every stage of the customer journey, governance frameworks recommended by experts, the full data breakdown by role, industry, and geography of respondents, and the CVS Health and LoadUp case studies in their full depth.

The full research is available directly from CX Network.

→ Access the full report: The State of AI in CX: From Assistive to Agentic

Frequently Asked Questions About AI in CX in 2026