In UX Research, we've long focused on making sense of mess. We turn chaotic threads and half-formed thoughts into clean narratives, polished summaries, and decision-ready insight. For a while, that was the work. That was the value.

But for the past few years, I've been rethinking that model. I've come to believe that the real value of research isn't in the output. It's in the infrastructure: the systems we design, the data we preserve, and the way we enable others to interact with that knowledge later.

Recently, my team ran an experiment that reinforced this belief and showed just how urgent this shift really is.

The Setup: Turning an Archive into a System

We'd spent years running a weekly insights call with client-facing teams. Sales, relationship managers, product folks would come together to share what they were hearing from the market. Every week, we captured that input in short internal reports. Every quarter, we rolled them up. Over time, it became a deep archive of qualitative signal: disconnected, under-leveraged, but full of potential.

We pulled all that together into a single 99-page document. Then, as a stress test we fed that document into a GPT. We used ChatGPT's custom instructions, gave it the full archive as reference material, and asked it to act as a research assistant. The goal wasn't just convenience. It was to validate a theory: if the right data is preserved and structured, we don't need to guess what future questions will be. We can design for the unknown and answer them later.

And it worked. Ask the GPT things like "What did clients say about onboarding pain points in 2023?" or "Were there recurring asks related to workflow visibility?" and it returned structured, specific, on-demand insight instantly.

But the real power wasn't in the answers. It was in what the experience revealed.

Value Is Shifting, and We Need to Design for That

People don't want to read 99-page reports. They want to ask questions. They want to explore, orient, and find context when they need it, not when we decide to summarize it.

This wasn't new information to me, but watching teams light up when they could interact with research this way confirmed what I'd already been advocating internally. Synthesis doesn't need to happen up front anymore. It can happen later, in context, on demand, if we've designed our systems accordingly.

And that changes how we define value. The value of UXR no longer lives solely in polished outputs. It lives in what we preserve, in how accessible and queryable our raw input is, in the infrastructure we build for future insight.

The report becomes disposable. The data becomes the asset.

UXR's Role Is Changing, and That's a Good Thing

This is the shift I've been helping our team navigate: from insight producer to insight architect. From being a bottleneck to enabling fast, scalable, safe access to raw signal without everything needing to flow through a single researcher.

We've already moved in this direction. We implemented a rapid feedback framework that compressed our typical validation cycle from 30 days down to under 30 hours by streamlining how we collect and synthesize input from stakeholders. We developed a monthly testing process that's smoother, better documented, and less dependent on individual heroics to get insights into the hands of product teams. We've built internal tooling to reduce friction and create reuse.

But this GPT experiment was a helpful proof point. It was a tangible, working model that helped others feel the shift I'd been pointing to. It gave people access to insight without a meeting, without a deck, without a wait. And that's the point.

The Real Risk? Ignoring This Shift

The biggest risk here isn't that AI gives bad answers. It's that we keep designing research as if AI doesn't exist. That we keep optimizing for summaries instead of preserving the inputs those summaries come from.

If we were running those calls today, we'd capture them differently. We'd design the data exhaust from the beginning, not just so someone could write a report, but so someone else, months later, could ask a question we haven't even imagined yet.

That's the shift. From "what do I need to say?" to "what might someone need to ask?"

Designing for the Unknown

We are not going back and rebuilding the past. That's not the goal. The GPT we built isn't about applying old insight to new projects. It's about learning how to build forward. It's a glimpse of the kind of infrastructure we need, not to answer today's questions, but tomorrow's.

That means capturing raw qualitative input in structured, responsible ways. It means treating preservation, not polish, as the strategic priority. It means rethinking how we store and index research and designing for AI-assisted querying from day one.

It means recognizing that research is no longer a finished product. It's a living system.

Final Thought

The future of UXR isn't just delivering insights. It's designing systems that make those insights accessible, resilient, and scalable, even to questions we don't know yet.

This is the work I've been moving toward for a while. The GPT didn't change my thinking. It just helped prove the point. And now that we can see where this is headed, we have a responsibility to lead.