How do you know your AI is right and how often do you check?
As AI agents take on a larger role in customer service, that question is becoming harder to avoid. Companies may know how quickly an AI agent resolves an interaction or whether an issue is resolved without being transferred to a human agent. But those measures don’t necessarily reveal whether the agent gave the customer accurate information or whether anyone would notice if it didn’t.
New research from Metric Sherpa and Parloa suggests that AI deployment in enterprises may be advancing faster than the adoption of measurement practices needed to support it.
Confidence grows with AI maturity
The 2026 CX Observability Study surveyed 531 customer experience and contact center professionals responsible for CX measurement, along with 238 executives who set CX strategy and own its outcomes.
Among operators, confidence that current metrics reflect the real customer experience rises sharply as organizations mature their use of AI. The proportion who say they’re very confident increases from 38.0% at the piloting stage to 76.9% at the fully embedded stage, respectively.
Executives show a similar pattern. When asked whether they have enough visibility to make confident decisions, 52.0% of executives at organizations piloting AI say they always or often do. At the fully embedded stage, that figure rises to 92.1%.
While this data paints a clear picture that the more established AI becomes, the more confident organizations feel about their ability to evaluate it, the rest of the findings blur the artwork.
The blind spots aren’t disappearing
If confidence reflects a more mature measurement practice, organizations further along in their AI deployments should report fewer gaps in what’s being monitored. But that’s not what the study finds.
Operators report measurement gaps at roughly the same rate regardless of how much AI their organizations use. Executives also continue to express significant concern about undetected AI errors across adoption stages.
Accuracy is the least-measured AI risk
Among the 367 operators whose organizations use AI, 78.5% report having no metric for hallucinations or factual accuracy. Even among operators who say they’re very confident in their AI, 76.6% report the same missing metric.
The gap shows at the executive level as well. Only 27.9% of executives say hallucination or factual-accuracy data reaches them, the lowest rate for any safety or compliance measure included in the study.
This gap brings out two key risks in a company’s AI-powered CX strategy. The first is legal and procurement risk: if you can't show how an AI agent performs, you can't show that it complies when an auditor comes to check. The second is a risk to the customer experience itself: A wrong answer from an AI agent that nobody detects will keep reaching people and eroding trust and loyalty every time.
CX observability is a desired tool
The study defines CX observability as complete visibility into every customer interaction, whether a human agent, an AI agent or both handled it; so teams can monitor, understand and act on performance and compliance issues. In practice, CX observability gives teams a way to answer a few basic questions before they expand an AI agent's role: Which interactions does the AI handle? How is factual accuracy measured? How quickly are errors detected, assigned and corrected? Who can pause further expansion if the evidence doesn't exist? Does accuracy data reach the executives responsible for CX outcomes?
The term is still new to most respondents. Only 32.0% of operators and 30.6% of executives said they were very or extremely familiar with it. But once they read the definition, 56.8% of operators and 52.8% of executives said the approach would be very or extremely valuable.
As AI takes on more customer interactions, leaders and operators recognize that they need to be able to show how it performs, where their measurement gaps remain and who is responsible for addressing them. They just didn’t have a term to describe it.
The 2026 CX Observability Study explores each of these gaps in further detail, including the disconnects between operators and executives and the organizational barriers that keep insights from becoming action. Read the full report to find your own organization's CX blind spots, assess your current measurement practice with the five-part CX Observability Index and learn where to focus next.
