revenue leaves a business in three ways, and only one of them makes any noise. Some customers stay on the books but quietly disengage, buying less, opening fewer emails, renewing out of inertia rather than loyalty. Some complain loudly enough that someone in the company finally notices. And some simply stop showing up, with no ticket filed and no exit survey completed, so the business never learns why. Qualtrics used all three as the spine of its keynote at the Utah Jazz training facility in Salt Lake City on September 9, calling them wanderers, loud levers, and ghosts. I think that framing gets at something a lot of companies still don't act on: every one of those three groups represents money the business already has, or already had, and experience is the thing deciding whether it keeps it.

The Three Ways Revenue Walks Out The Door

Most experience programs are built to hear the loud levers, because complaints are the easiest signal to collect and the easiest to justify budget around. Wanderers and ghosts are harder, because neither one files a ticket. A wanderer still renews, still logs in, still technically counts as a retained customer on a dashboard, even as their engagement quietly declines toward the exit. A ghost doesn't even give you that much warning; the relationship just ends, and the first sign is a churn number you see weeks later.

Qualtrics is proposing that its XM Data & AI Platform can surface all three cohorts before they turn into a lost renewal, using the same underlying data and the same three capabilities: simulation, prediction, and what the company is calling trusted outcomes.

What Simulation, Prediction, and Trusted Outcomes Actually Do

Simulation builds a synthetic version of a customer population, so an organization can test a pricing change, a policy shift, or a new offer against a digital twin before it reaches a real market. Qualtrics says these are fine-tuned on its own research corpus rather than general-purpose models improvising a persona, which is the distinction that matters if you're deciding whether to trust the output.

Prediction narrows that from a population to a person. Where simulation tells you how a segment is likely to behave, Qualtrics' Experience Prediction Models are built to estimate what a specific customer is likely to do next, while there's still time to change the outcome.

Trusted outcomes is the action layer. Before anything reaches a customer or employee, it's checked against the organization's own rules, and Qualtrics is explicit that customer data stays inside that organization's environment rather than training a shared model. In the demo Qualtrics ran, a fictional hotel brand used the same three-step chain, simulation to identify a lapsed cohort, prediction to individualize the offer, and automated action across marketing, an AI concierge, and the property system, to close a loop that would otherwise have stayed open.

Why This Isn't Guesswork With Better Marketing

The common mistake, in my experience, is treating experience as a customer service line item, something you fund modestly and hope pays off in goodwill. Companies that have actually orchestrated a good experience, across product, support, and every other touchpoint, tend to win, and there's no shortage of examples proving it. The gap Qualtrics is naming is the distance between what a customer expects, what they actually get, and what a company should be doing about that difference in the moment it happens.

What makes this different from a person's intuition about what "feels right" is the volume and combination of signals behind it. Instead of throwing every possible action at a customer and watching what sticks, the platform is built to combine feedback, behavior, and context into a prediction of what will actually move an outcome, then act on it inside whatever system the customer is already touching, rather than a dashboard nobody in the flow of work ever opens.

That's also where the experience ontology matters more than the AI itself. Qualtrics describes it as the layer that gives a signal meaning: what it is, whose it is, where it sits in the relationship, and what the next best action should be. Without that context, a model produces answers that sound plausible and are frequently wrong. With it, the same model has something to reason from.

The Data Behind The Claim

Qualtrics' pitch rests on the size and specificity of what it's feeding the models. The base dataset spans more than two decades and 18,000 organizations. The $6.75 billion acquisition of Press Ganey Forsta, closed in May 2026, adds four decades of healthcare experience data from more than 41,000 facilities, along with the regulatory governance that comes with handling patient information, which Qualtrics is applying across the rest of its experience data as a rigor standard.

On results, Qualtrics pointed to TruGreen, which deployed Experience Agents across millions of customer interactions and reported $7 million in ROI from closed-loop feedback, a 30 percent reduction in customer escalations, and $30 million in total ROI once digital optimization and retention were included. SiriusXM's Steve Lilly, Director of Customer Experience, framed the appeal in terms of speed: the organizations that win are the ones that can hear a signal across any touchpoint and act on it in real time, not after a quarterly report surfaces it.

What Happens Next

• The XM Data & AI Platform itself isn't shipping until 2027. The X-Loop Diagnostic Service and the new Reputation Experience capability are available now, which means the near-term adoption signal to watch is whether organizations buy into the diagnostic groundwork before the platform exists to act on it.
• Qualtrics addressed pricing directly at the keynote: the new capabilities are bundled into the existing XM Suite rather than sold as a separate line item, which matters for a platform positioned to replace several existing tools rather than add one more to the stack. What wasn't detailed is how usage-heavy features like simulation and prediction get metered once the platform actually ships.
• The TruGreen and SiriusXM proof points are early and vendor-selected. Whether the prediction layer holds up across industries with thinner data histories than 18,000 organizations is the real test still ahead.