When I last sat down with Jason Maynard back in February, he was just a few hours into his new tenure as Qualtrics CEO, replacing Zig Serafin. It was becoming clear at that time that Qualtrics was about to go through a significant change – helping its customers move from listening and understandingce’. At the time, Maynard told diginomica:
Experience Management is more important than ever, especially with AI. The key thing is, when you think about human experience and understanding human experience, you have to act on what matters in context. That matters massively when you’re trying to deliver an agentic experience.
The thinking is that as services, processes and work are increasingly carried out by agentic AI, the only way to really understand what is landing with employees and customers is through additional experience context. In other words, not just using context to understand why a transaction took place, but understanding how people felt – or will feel – about it.
Qualtrics’ changing strategy aligns with a broader shift in the enterprise software industry. As AI agents take on more work, vendors are being forced to rethink where their value lies – what context agents need, which systems they act on, who owns governance, and whether people need to log into applications at all. If experience data is going to be in the mix for enterprise buyers, it has to show up and provide value where agents and people are actually making decisions. ‘After the fact’ understanding is less helpful as systems become increasingly autonomous.
A few weeks after Qualtrics previewed its latest XM Data and AI platform, which speaks to all of this, Maynard sat down with diginomica to explain his thinking.
From insights to outcomes
But first it’s helpful to have some background. In our February interview, Maynard positioned Qualtrics as a ‘system of decision’. At X4 in March, the CEO went further and said that using survey data to understand customers or employees is a relatively established problem – the harder piece is acting on that understanding in the moment, before a frustrated customer or employee walks away, rather than working out what went wrong after the fact. Knowing why something went wrong, after it went wrong, is an “autopsy”, he said.
In September, Qualtrics unveiled its new XM Data & AI Platform, which is due in 2027. This is built around three capabilities to tackle this ‘in the moment’ problem: simulation (testing how a decision lands before committing to it), prediction (using small language models for forecasting what a specific person is likely to do), and trusted outcomes (checking any action against an organization’s rules before it reaches a real person). Qualtrics also says customer data should never leave the organization’s environment or train shared models.
This also brings us to a word that has dominated the industry’s agentic conversations this year: ‘headless’. A number of large platform vendors have positioned their systems essentially as back-ends for AI agents and other interfaces to consume, rather than screens or applications to log into. If the work increasingly happens elsewhere
However, Maynard isn’t a fan. He said:
I’ve been making fun of the word “headless”. You could also call it “brainless”.
That’s not because Maynard is wedded to the old approach of screens and consuming insights for who controls the customer relationship
Headless is brainless
The ‘headless’ discussion is important because it goes back to a question I’ve been pondering since Maynard took over: how does experience insight actually play out operationally, across service, CRM and other systems that will have their own agents? In response, Maynard said:
My agent will talk to your agent, is the glib short answer.
His longer response is that he expects the bulk of what Qualtrics produces to stop living in dashboards:
In some number of quarters, years, the shift in our business will be that more of what we inference – the insight we inference – will end up embedded in something other than a dashboard. I’ve told folks in meetings that at some point, 90, 95% of what we do will be embedded insights. That’s our plan.
Whilst this might sound familiar, it’s important not to confuse this with what other vendors are talking about – this is not headless in the same way others are approaching it. The difference, for Maynard, is who decides where the insight goes and what gets handed over in the process. For example, he pointed to personal AI agents that are growing in popularity, but are attempting to sit between brands and their customers. The brands are already blocking them, but this points to the problem Maynard and Qualtrics are trying to solve:
If you buy software from a company that says “we’re headless”, it implies you get access to everything that’s going on – you just pipe it in, suck it all out. When I hear that term headless, that’s why I use the word brainless with it – because if the head’s gone, so is the brain. Words have meaning, as far as I’m concerned.
I ban certain words in the company. I’m like, do not ever put the word headless on any slide, because that is not our product strategy. That is not our vision. Our vision is an embedded insight, based on a simulation of what we think might happen – or what the organization or the data science thinks might happen – and then we create a prediction model that generates an insight token.
What does this mean in practice? Until recently, the approach would be to send a survey after an interaction, apply those responses up into an NPS or CSAT score, and hope that someone acts on the results of that weeks later and brings about some change. With the new platform, an organization would be able to simulate how a price change lands with its best customers before committing to it, predict which customers are likely to react badly, and check any response against its own rules and policies – acting in the moment, rather than to a report weeks later.
The insight token would get embedded wherever the customer wants it, including third party agents if that’s what they desire. Other vendors are also leading heavily on governance and context too, within their headless architectures – but the difference for Maynard is what actually gets handed over. My understanding is that Maynard sees a headless system as opening up its data and logic for outside agents to pull from. Whereas, the way I understand it, Qualtrics wants to keep the data, and the reasoning applied to it, inside its own platform, and send out only the finished insight – one checked against the organization’s own rules.
Two things can be true at once
This argument for giving customers control runs through a lot of what Maynard is talking about – and this is underpinned by the conversations he’s had with customers over the past nine months. On the one hand, he told diginomica that customers are experiencing a pace of change that is accelerating faster than anyone expected:
No matter what happens, you get an experience with one organization – and even in your consumer life, you get a great experience – and you assume everything else you touch should be just like that. So you no longer compare vendor A to vendor B in category X. You compare vendor A in category X to every other category in the alphabet, and you’re like, “Yeah, but why can’t it be like that?”
On the other hand, which may seem counterintuitive, customers are also trusting AI less and want more control over their data:
Everyone is looking at this now from a PII perspective, more than they did even 12 months ago. The notion of sovereign AI – two years ago, everybody was like, “What does that mean? That’s just for defense agencies or intelligence communities.” Now it’s, “No, we have to maintain control and protection of our data,” which increasingly is viewed as the IP.
This tension is creating what Maynard calls “cross-currents” and he said that both things can be true at once: customers want to adopt AI at speed to meet the changing expectations of their customers, but they also want control and want to be able to trust what they’re using.
Trusted, not just outcomes
Maynard went on to say that whilst Large Language Models are powerful, most enterprise interactions are governed by rules that aren’t up for debate. Other software vendors are also becoming more pragmatic about deterministic rules working alongside probabilistic models. For instance, Maynard said, an airline loyalty program deciding who gets an upgrade will use established rules inside an organization. He said:
You can’t do this without understanding the deterministic side, the business rules in effect in the organization. And then you get into the question of, how do you distill those rules? Do you then distill them and embed them in somebody else’s harness that lives with the frontier lab? I’d say most organizations don’t want that, because they’re like, “Well, that’s kind of our IP. That’s kind of important stuff. Let’s keep that in our own harness.”
What’s interesting about what Qualtrics is proposing is recognizing the value of AI, but it isn’t looking to throw the baby out with the bath water. Organizations want to still own IP, they want to know how their data is used, and they want to couple AI with rules.
Again, a stickler for semantics, this is why Maynard talks about ‘trusted outcomes’, rather than just outcomes. Qualtrics has long described what it does in single words – listen, understand, act. And two new words were added with the new platform: simulate and predict. As such, it pained Maynard to include a two-word phrase in the marketing materials -’trusted outcomes’. But he said it was necessary:
If I don’t put “trusted outcome” there and just put “outcome”, it makes no sense. And if I just put “trust”, it didn’t make sense. It’s those two things tied together.
Every organization, government or commercial, cannot be in the business of serving slop. Brainless, headless slop. Your business, done.
A new set of buyers
As I noted at Qualtrics’ event in March earlier this year, if experience data is to play a crucial role in cross-enterprise AI decision making, this requires that the vendor is thinking differently about how it sells its platform and how it structures its go-to-market. Qualtrics needs to sell to a different persona in the enterprise, not just HR or CX leads. After 15 or so years in which every line-of-business function bought its own SaaS applications, organizations are pulling things back together. He said:
I was talking to a COO on Monday, and they’re like, “We used to have like 3,000 to 4,000 applications.” I’m like, “3,000? 4,000?” He’s like, “Nobody knew, man. Nobody knew. What I can tell you is we’re down to 800.”
So what’s happened is, when Qualtrics started, like every other company, we had our ICP, and we’d sell to that person. In almost every organization now, it’s getting uplifted, and the CIO or Chief Data Officer or Chief AI Officer is in the group chat, if you will.
Maynard added that some organizations are restructuring around hybrid roles that look across functions – “the right org chart candidly”. Qualtrics has responded to this, according to Maynard:
We’ve added senior go-to-market sales leadership who have lived in organizations where they’ve called higher up. We still take care of and serve our constituents who are the functional owners, but those functional owners are now saying, “Well, you’ve got to go talk to so-and-so anyway.” Because AI is only going to work when you reason across the silos.
That last line is the crux of it. Customers are struggling to find value from agentic AI right now, and I’d argue that’s unlikely to change until they rethink their organizational structures. A vendor whose pitch rests on reasoning across silos needs to be in the room with the people who own all of them.
My take
Throughout my coverage of Qualtrics this year, I’ve maintained that its advantage, if it can execute effectively, is that it has horizontal opportunities across the enterprise. It sits across operational systems with data about how people actually feel about them. This also speaks to another concern I’ve had about agentic systems: if everything is frictionless, what do you compete on? Qualtrics brings something to the table that’s different here, it helps organizations compete with experience data. That’s potentially very valuable, if it can help customers understand why that’s important and also help them change what they do as a result.
The customer stories we’ve looked at this year show what this could look like on the ground. TruGreen used Experience Agents to cut escalations by 30% and lift retention by 8 per cent. Trillium Health Partners moved from paper surveys to patient feedback managers could act on within 24 hours, rather than three months. The PGA Tour brought siloed feedback from fans, employees, players and partners into a single view. In each case, the value came from experience data showing up where decisions were being made.
I also think Maynard has a good point about language. Semantics matter here, particularly when agentic AI brings so much change and customers are concerned about what it means for their organizations, people, customers and data. Qualtrics will place its insight wherever the customer needs it, including a third-party agent – but the customer keeps control of its data, its rules and, ultimately, its relationship with the people it serves. Given the trust concerns he’s hearing, that framing is likely to resonate with many buyers.
Maynard also shared some interesting developments in the pipeline that I agreed to keep off the record for now. What I can say is that they reinforce the direction set out here: this is a company aiming to go well beyond its origins in surveys and sentiment, and to put experience data front and center across the enterprise as decisions are being made. I’ll be reporting back on that soon.
