AI is reshaping service delivery, and with it the metrics providers use to prove the investment worked. Many are positioning these metrics as experience measures. Most of them measure something else entirely.
Experience Level Agreements (XLAs) are frameworks for defining and measuring experience outcomes associated with technology services, sometimes incorporated into formal service agreements. They measure KPIs associated with the end user experience, and they differ from SLAs, which are strictly operational and transactional in nature. The so-called “Watermelon Effect” has long highlighted situations where operational SLAs appear healthy on the surface, but employees remain dissatisfied with their technology experience. XLAs attempt to quantify the human aspect of technology performance.
Experience was traditionally measured through satisfaction surveys, feedback on closed tickets, sentiment analysis, and “day-in-the-life” shadowing. As Digital Employee Experience (DEX) tools became common, endpoint telemetry and desktop analytics began feeding those measures directly.
Why AI Metrics are not XLAs
Generative and agentic AI make it possible to deliver uninterrupted and seamless technology experiences. Generative AI summarizes and consolidates information. Agentic AI introduces autonomous workers that act without a person initiating the request. Both produce a new class of metrics: incidents resolved by AI versus humans, AI utilization, AI agent effectiveness, Copilot adoption, AI PC device provisioning, etc. Many service providers will position these as the next generation of XLAs. They’re not.
End users generally do not care whether an issue was resolved by a human technician or an autonomous AI agent. They care about getting back to work quickly and correctly. AI utilization and automation rates describe operational effectiveness, not end-user experience.
Consider a scenario where a service desk agent resolves an IT issue in five minutes, and another scenario where an AI agent does the same in 30 seconds. If the issue is resolved correctly and the employee is back at work, the level of satisfaction is roughly the same. Nobody notices the utilization rate, model accuracy, or automation rate behind either one.
AI changes the delivery model, while the experience outcomes remain fundamentally human. They are evaluating effort, friction, productivity and confidence. Hence outcome matters, resolver does not.
Where AI can really change XLAs: Persona-based Experience
The most useful aspect of AI in experience measurement has little to do with efficiency. It’s the ability to build persona-specific intelligence into the measurement itself. Thirty seconds versus five minutes may not matter much for a knowledge worker. For a clinician or trader, it can be the difference between a working day and a failed one.
Historically, workplace services have treated employees the same way with a standard laptop, refresh cycle, and standard support model. AI makes it possible to define personas around what people need from technology.
A financial analyst working in spreadsheets and reporting tools may get real value from Copilot without requiring a high-end AI workstation. Similarly, a contact center agent needs fast response times and conversational assistance, but not local AI processing. A data scientist needs AI acceleration, local inference capabilities and significantly higher compute resources. Having the same device strategy for these three distinct personas make little sense, and neither does one set of XLAs.
If XLAs measure whether technology enables someone to achieve the desired outcome, then the device itself cannot be treated as a standardized asset independent of its user. Persona intelligence is not only a better way to measure satisfaction. It’s a better way to measure cost. Five minutes lost by a financial analyst and five minutes lost by a contact center agent are not equivalent losses to the business, and a measurement model that treats them identically cannot tell an organization where its capacity is going.
Persona-intelligence defines AI-device usage
With AI-powered devices, the market conversation about device lifecycle management centers on hardware specifications such as NPU performance, TOPS, battery efficiency and Copilot readiness. The more useful question is which personas get a better experience from an AI PC, because not all of them will, and the ones that do won’t generate the same value from the capability.
That changes what an XLA should measure. Time taken to deploy an AI PC tells you almost nothing. Whether the right employee received the right AI tools at the right time tells you something more meaningful.
The same shift also challenges the traditional asset management approach. Device lifecycle management decisions have long focused on device refresh cycles, recycling, asset recovery and refurbishment. AI introduces a variable those frameworks weren’t built for, which is how much value the device generates for the person using it.
Extending the life of a device is sometimes the right call. Upgrading a knowledge worker to an AI PC is sometimes worth considerably more than the savings from keeping an older machine in service. This is not just an economic or environmental decision. The objective is to maximize employee outcomes while balancing cost and sustainability.
What should the next generation of XLAs measure?
The next generation of XLAs should focus on the following questions:
- How quickly does an employee return to productive work?
- How much effort is required to accomplish a task?
- How often does technology interrupt work, and how much of that interruption was avoidable?
- Does the employee have the right device for their persona?
- Does the employee trust workplace technology?
The future of XLAs remains human-centric. AI will transform how experiences are delivered, measured, predicted and personalized, and it will make experience management more proactive and specific to the personas. But employees don’t experience AI. They experience whether the technology enabled them to do their work. That is what XLAs have always been about. What changes is our ability to define what better means for each of them.
Pomeroy builds persona-centric experience models for enterprise workplace environments. Our advisory practice develops experience indicators by persona and industry vertical, mapped to the business outcomes those personas drive. Learn more about our advisory and consulting services.
