In March, a group of banks led by JPMorgan Chase tried to syndicate $5.3 billion of debt for Qualtrics . The proceeds were earmarked for the company's $6.75 billion acquisition of Press Ganey Forsta . The deal would have been the largest consolidation in the Voice of the Customer category — putting Qualtrics, Press Ganey, Forsta, and InMoment (which Press Ganey had already acquired) under a single owner. Medallia and Sprinklr would have remained the other major independent players Eleven banks had already committed. By the standards of investment-grade leveraged finance, the deal was conventional.

It failed.

Bloomberg reported it in plain language. Investors in the loan and bond markets declined to fund the deal, citing AI disruption risk to the survey software business as their main concern. Qualtrics's existing $1.5 billion loan, which had been trading near par, fell to roughly 86 cents on the dollar.

This is unusual. Bond investors are not, as a class, paid to have aesthetic preferences about technology categories. They are paid to be right about whether a borrower can service its debt over the life of the instrument. When eleven of them, after committing capital, walk away from a deal this size and give a single reason, the message is clear. They don't believe the cash flows behind that business will be there to repay the loan.

Six weeks later, the same message arrived from a different direction. Thoma Bravo handed Medallia to its creditors in a debt-for-equity swap that wiped out most of the equity paid for the company in 2021 — the resolution of a debt position I've covered at length elsewhere. The bond market's verdict on Qualtrics in March, and the lenders' verdict on Medallia in April, are the same verdict delivered through different instruments.

For enterprise CX leaders sitting on multi-year contracts with Qualtrics, Medallia, or Sprinklr, this is worth a moment of attention. The capital markets and your CX practitioners are, through entirely different mechanisms, arriving at the same conclusion.

What the platforms' own research is saying

Medallia's 2026 State of Customer Experience Report is, in a quiet way, one of the more remarkable documents the industry has produced this year. The report draws on 552 CX practitioners and 1,522 consumers. The benchmark sample is more than 600 enterprise programs running on the platform that published it.

The headline finding is a 49-percentage-point gap between practitioners and consumers on the question of whether CX is improving. Sixty-six percent of practitioners say it is. Seventeen percent of consumers agree. A gap of that size is not noise. It is not measurement error. It is what happens when an industry measures itself by talking to a small, self-selecting fraction of the people whose opinions actually matter to the business. Medallia put it on page one.

Three further numbers from the same report are worth holding together. Survey response rates fell 11% year over year. One in three departments getting CX insights takes no action on them. And 75% of practitioners now agree that surveys alone are not enough — even as surveys remain the primary input to the platforms they're using.

This is the published research of a category leader. The Gartner Magic Quadrant for VoC Platforms, released a few weeks earlier, reaches the same conclusion from independent analysis. Most CX programs have not kept up with their vendors' AI advances and remain largely survey-driven. Gartner identifies three gaps across all category Leaders: a missing link to financial outcomes, no analytics for the customers who don't respond, and patchy adoption of the AI features the vendors have already built. None of the three is on a near-term roadmap.

You could think of it this way. When the analyst community, the platforms' own customers, and the bond market all reach the same verdict in the same quarter, it usually means something.

The architectural problem AI cannot fix

The leading VoC platforms have made real AI investments. Qualtrics has Experience Agents and an Insights Explorer. Medallia has Athena. Sprinklr has trained models on more than 100 million CX data points across thirty-plus channels. These are not vapor capabilities. They produce real value for organizations managing high-volume consumer feedback. They do not, on the other hand, fix the problem of customers who never gave feedback in the first place.

The constraint is architectural. The AI sits downstream of feedback collection. It processes, summarizes, and draws insight from signals that have already arrived. Customers who do not respond to surveys, do not write reviews, and do not post about you on social platforms generate no data for these models to work with.

In most enterprise programs, that is the majority of the customer base. Gartner notes that typical VoC programs hear from fewer than 30% of their customers. In digital-first sectors, where survey fatigue is acute, the figure is often lower.

The customers who matter most to retention are often the ones generating no survey signal. The client who interacts heavily but never responds. The account whose engagement patterns are quietly shifting. The relationship that will not renew. These are not the customers a survey-based system sees clearly. Adding a more sophisticated AI engine to survey data does not expand its coverage. It optimizes the analysis of a signal that was incomplete to begin with.

There is a real distinction between two things that often get conflated. The first is using AI to make survey processing faster and smarter — better summarization, better routing, natural-language search across existing feedback. That is a genuine efficiency gain. The second is using AI to build continuous customer understanding from behavioral and operational data, whether or not anyone responds to a survey. That is a different capability altogether. Gartner identifies the second as the gap the Leaders have not closed. It cannot be closed by adding AI to the first. The Leaders have, in general, chosen to add AI to the first.

The latest example arrived in May. Qualtrics's flagship announcement at its X4 conference was Synthetic Panels: large language models that simulate how consumers would respond to a research question, in hours rather than weeks, at half the cost of an actual panel. The platform built on listening to customers has decided, in the face of declining response rates, to simulate them instead. It is, in its way, a clarifying answer to the architectural problem. Adding AI to a survey-based architecture does not fix the coverage gap. It produces a substitute for the data that isn't there.

Why incumbents almost never adapt

There is an obvious counterargument to all of this, which is that the incumbents will simply build the missing capability themselves. They have the customer relationships, the data, the engineering teams, and — at least nominally — the capital. Why would the new architecture not come from them?

The historical record is the first answer. Mature software companies that innovate their way out of a maturing technology base are remarkably rare. Oracle is one of the few — but Oracle's escape was financed by serial acquisitions of newer technology, not by organic invention. Buying your way out is a different game from innovating your way out. And it requires capital that the incumbent survey platforms now appear unable to access on reasonable terms. The Qualtrics bond deal failing, and the Medallia debt-for-equity swap that followed, are the closing of that escape route.

Geoffrey Moore's analysis of incumbent zones gets at part of the structural reason. Public companies under pressure to grow profits struggle to make the long-horizon R&D investments that genuinely transformative technology requires. The capital discipline that makes a company attractive to investors during the harvesting phase is the same discipline that stops it funding the next architectural shift. Startups, by contrast, can run capital deficits because they have nothing yet to harvest.

But the capital problem is, in my view, not the deepest one. Two other forces work against the incumbent.

The first is the weight of existing code. Mature enterprise software platforms typically spend something on the order of 80% of their engineering capacity on maintenance — keeping what already exists working, integrated, secure, and compliant. Whatever is left funds the next thing. When the next thing is incremental, that is enough. When the next thing is a different architecture entirely, it is not.

The second force is more interesting and more under-appreciated. The incumbent's customers bought the old proposition. They did not buy the new one. They were sold a vision of being the world's greatest survey-powered analytics platform, and they invested heavily in that vision. Walking into the customer briefing the next day to explain that surveys are not, after all, the future is the kind of conversation you do not have with a spouse if you would like to stay married. Customers have a nasty habit of looking around when they sense a technology transition is on the horizon. The rational vendor strategy is to make sure they do not start looking.

Which is why the only viable position available to the incumbents is to argue that AI is incremental, not transformative. The story has to be that AI doesn't change the foundation of what they're doing — it just makes their current technology better. If that turns out to be the right read, they are in excellent shape. AI becomes another toolkit that produces more polished insight from the data they were already collecting, and the customer relationship continues uninterrupted.

The trouble is that the incremental story works for some technology transitions and not for others. In the case of survey data, it does not work. Adding AI to survey data does not solve the underlying problem with survey data. It makes the insights drawn from a signal that is incomplete to begin with look more polished, which is not the same as making them more accurate. The category Leaders are, by virtue of their installed base, locked into telling the story that the underlying evidence is least equipped to support.

In Medallia's case, the story now has to be told to a customer base that just watched its previous owner exit. The new owners inherit the operating conditions that produced the outcome. You can argue Medallia doesn't deserve any of that. It doesn't change them.

The reallocation argument

The temptation, when an industry receives signals like these, is to read them as a call for wholesale replacement. The evidence does not support that, and neither does common sense. CX program infrastructure — embedded workflows, frontline culture, executive alignment, historical benchmarks — has real value, and that value was not created cheaply. It is worth preserving.

What the evidence does support is a deliberate repricing. Survey platform contracts were set when the category had limited competition and surveys were the only way to listen to customers systematically at scale. Neither condition holds in 2026. Response rates are falling. Capital markets are marking down the future cash flows of survey-centric businesses. And a new generation of tools now exists that can cover the full customer base, drawing on behavioral and operational data, without requiring a survey response. Those tools connect directly to revenue outcomes.

Recent research has shown that behavioral data alone is enough to predict churn with strong accuracy. Login frequency. Feature adoption. Support ticket velocity. Engagement trend lines. No survey responses required. The question that matters most to revenue protection — which customers are at risk, and when — can now be answered through a better data setup than the one most CX programs are funding.

The reallocation is not "stop surveying." It is to stop paying premium prices for what is now a commodity layer of the stack. The savings go toward the predictive analytics layer — the layer that the Gartner analysis, the Medallia self-assessment, and independent research all identify as the missing capability.

This is reinforced by what's happening to the broader enterprise AI budget. PwC's 2026 CEO Survey of 4,454 executives across 95 countries found that companies applying AI widely — to products, services, and customer experiences — earned roughly four percentage points higher profit margins than those that didn't. The gap is widening rapidly. The firms capturing that margin are not the ones that bolted AI features onto existing workflows. They are the ones that redesigned their analytical architectures around AI from the start. A separate TechCrunch survey of enterprise venture investors put it more bluntly: budgets will rise for the narrow set of AI products that clearly deliver, and decline sharply for everything else. The Qualtrics bond deal failing is, in part, the bond market pricing in that prediction.

What to ask before the next renewal

A handful of questions are worth putting to any incumbent platform before the next contract review. The vendor will be ready for some of them.

What share of our customer base does this platform actually generate insight for? And how is that share changing as response rates fall? If the answer is bounded by survey respondents, an architecture that depends on response is, year after year, less representative of the full client population.

Can this platform demonstrate — in production, not in a sales deck — a connection between CX signals and financial outcomes such as retention revenue and renewal probability? Gartner has flagged this as the critical missing capability across all three category Leaders. The test is whether the platform produces that connection routinely as a standard output, or describes a future in which it might.

What is the fully-loaded cost of ownership at year two and year three? That number should include professional services, managed services, and configuration support. Gartner's own assessment of Medallia notes that TCO is higher than peers due to complexity and the frequent need for managed services, with implementation times the longest in the category. A platform that needs heavy ongoing vendor involvement to run is not delivering self-service analytics at the price of self-service analytics.

If one in three departments ignores CX insights, is that really an adoption problem? Or a symptom of insights that aren't connected enough to operational decisions? Adoption programs have a role. But persistent low action rates, across very different kinds of organizations, more likely reflect a property of the insight itself. Too aggregate. Too backward-looking. Too disconnected from the actual decisions operational leaders are making.

And finally, what is the contingency if the Qualtrics consolidation brings roadmap disruption or pricing renegotiation in the next eighteen months, and if the Medallia restructuring leads to product or service interruptions? Gartner explicitly flagged the Qualtrics consolidation as a near-term concern. The Medallia event has now added a second one. Knowing what alternative providers can do is prudent due diligence, not a distraction.

A structural transition, not a vendor preference

None of the evidence here says survey-based CX programs have failed, or that the platforms supporting them have not delivered value. They have. What it says is that a category built around the survey as the primary instrument of customer understanding is now meeting the limits of its design. Those limits are documented by the vendors themselves, confirmed by independent research, and increasingly reflected in how capital markets price the underlying businesses.

The transition from survey-centric to behavioral and predictive customer analytics is not a new idea in 2026. What is new is the convergence of signals making the case for acting on it now rather than deferring. Response rates are falling. The link to financial outcomes that boards and CFOs increasingly demand remains undelivered by the category's Leaders. In March, capital markets declined to finance $5.3 billion of debt for the sector's most prominent company. Six weeks later, the sector's second-largest player was handed to its creditors. The tools needed to close the gap — continuous, predictive insight for the entire customer base, drawn from behavioral data — are in production today, not on a roadmap.

The strategic question is not whether to keep the program investment. It is whether the platform contracts behind that investment are still priced to reflect their current value. And whether the analytical capability most needed to protect and grow customer revenue is being funded at the level the evidence now warrants.

The platforms that will define the next era of customer intelligence are not the ones that listen more efficiently. They are the ones that predict more accurately, across the full customer base, before decisions have already been made.

I'm Richard Owen, founder and CEO of OCX Cognition. We build predictive customer analytics for companies who'd prefer to know which customers are at risk before those customers have already decided to leave.