What’s the Future for UXR? Here's what My Crystal Ball Says

I consulted my crystal ball about the future of UX research.

It cost forty dollars. It shipped from a warehouse that also sells novelty ice cube trays. So far it has been right about exactly one thing, which is nothing.

But I dusted it off, I held it in both hands, I leaned in and asked it what becomes of us. It showed me a future. Then it buffered. Then it showed me a slightly different future. Then it asked me to rate my experience on a scale of one to five.

So treat all of this accordingly. None of it is forecasting. It is me squinting into a foggy orb and narrating my anxieties in a confident tone. Which, now that I write it down, is also the job description for most of LinkedIn.

Here is what it showed me.

Mixed methods wins, but backwards from how everyone says it

Everyone agrees on this one. That is usually the first sign a prediction is too comfortable to be the whole story.

Pure qual is cooked. Pure quant was always a little lonely. The future belongs to the person who can hold a transcript in one hand and a regression in the other. The quant bar rises from “I made a bar chart in the survey tool” to something closer to data science with the serial numbers filed off. You can write SQL without crying. You can pull your own data. You understand an experiment well enough to ruin a PM’s afternoon with the word “underpowered.”

So the obvious move, the one everyone reaches for, is to keep climbing. If a little quant got you hired, more of it gets you promoted. You learn SQL. Then regression. Then experimentation. Then a little Python, because at that point why stop. Each rung feels like progress, and eighteen months later you are doing a junior data scientist’s job, slightly worse than an actual junior data scientist, and filing it under research.

Call it data science lite. The diet version. This is where most of the field quietly thinks the future is. Up the quant ladder, one bullet point at a time, until a UX researcher is a data scientist who also runs the occasional interview.

Fine. The bar rising is real, and that part I think is right. But the conclusion almost everyone draws from it, that the move is to become more quantitative, is the flaw.

Watch what the tooling actually does to that logic.

If the machine drops the cost of quant to near zero, the smart move for a person is not to sprint toward the thing the machine now does for free. It is to run the other way, toward the part the machine is bad at. Framing the question. Knowing which finding the whole decision rests on. Sitting in a room full of executives who have already made up their minds and being the one who says the inconvenient thing out loud.

That is comparative advantage. It points away from quant, not toward it.

So the honest prediction is stranger than the popular one. Your output gets more quantitative, because the stack handles the numbers. You get more qualitative, because that is where the part nobody can replace actually lives. The thing that is “mixed methods” is you plus the tooling, fused into one slightly cursed cyborg. Not you on your own.

This is the trap hiding inside data science lite. It sounds great right up until it means you do just enough quant to be dangerous and not enough to be right. You lose the quant fight to the people who do quant for a living. You have already sold the house on the qual depth that made you worth hiring. Congratulations. You have arrived at the exact center of the field, the one spot where you lose to everybody standing at either edge.

There are a lot of people standing in that exact spot right now, looking pleased with themselves. I do not have the heart to tell them.

Then there is the merger, and that is where I get nervous

Here is the big one. Market research, UX research, and data science stop being three tribes with three Slack channels and three separate sets of feelings. They fold into one function.

We call it Research. Or Insights, if there is a CMO who likes the word Insights. Or Customer Intelligence, if somebody is angling for a promotion.

I believe the walls come down. I do. They were never principled walls to begin with. They were org-chart accidents, leftovers from whichever VP happened to be hiring in 2014. And the new agentic tooling does not care about your tribe at all. Feed it a survey, an interview, and a clickstream and it blends all three into one answer without checking whether the priesthood approves. The knowledge merges whether the org chart catches up or not.

But watch the other direction.

Because “they merge into one function” is one future. The opposite is just as alive. Research stops being a function at all and dissolves into the teams it serves, leaving a thin strategic core on top and a lot of PMs running their own studies with a tool that politely stops them from writing leading questions.

Which one happens is not an intellectual question. It is a political one. It comes down to a single thing. Does leadership see research as a moat or as a cost center.

Moat, and you get consolidate-and-elevate. The unified intelligence function. The version that ends up on a conference stage.

Cost center, and you get democratize-and-gut. The same merger, wearing a much sadder outfit.

I have to be straight here, because I have watched how these mergers arrive. They do not arrive because a VP woke up enlightened about unified intelligence. Look at the DoorDash reorg. Product and ops merged. Design and engineering merged. Hiring froze. That was not AI reinventing the org. That was cost rationalization after an acquisition, with AI handed the microphone to explain it in a flattering voice. The restructuring came first. The AI story was the press release.

So when the great Research merger lands, my money says it shows up the same way. Lean. Output first. Headcount math in an AI-efficiency costume. The “Research” function that walks out the far side is plausibly the hollow version, not the elevated one we keep drawing on slides.

I would love to be wrong. I do not think I am.

The middle falls out before anything else does

This is the prediction I enjoy least, which is how I know to trust it.

The field splits into two real jobs and the gap between them keeps widening until you can drive a reorg through it.

On one side, the senior researcher who owns framing and judgment and governance. The person companies still pay real money for, because the machine cannot be held accountable in a meeting.

On the other side, the research engineer. The one who builds the pipelines and the knowledge systems everyone else self-serves on.

And in the middle, the generalist whose whole identity was “I run studies.” That person gets automated from below, because the tooling runs the study now, and out-thought from above, because judgment did not commoditize.

None of this is new. It is just your turn. The cloud did it to the DBA. Digital did it to the photo lab. Streaming did it to the video store clerk who had strong opinions about the staff picks shelf. The middle of a profession is the first thing to go the moment the floor and the ceiling stop needing it to talk to each other.

Somewhere right now there is a LinkedIn post titled “5 Reasons Researchers Are More Essential Than Ever.” I have decided to read it as gallows humor.

The frame is the one thing AI cannot do for you, and it is about to need a paper trail

Here is the part I am most sure of, which is rare for a piece about the future.

Strip out everything AI is going to swallow. The data collection. The transcription. The first-pass analysis. The synthesis that used to take a quant UXR three days. Take all of it. What is left is the part that was always the actual job, and it already has a name.

It is the frame. The organization’s accumulated, actively maintained model of its users. Not the archive of studies you have run, that is a repository, and the repository is exactly the kind of thing AI is delighted to manage for you. The frame is what the org currently believes about who its users are and what they are trying to do, and whether that belief is still true. I wrote a whole piece on it.

That is the thing that does not commoditize. A machine can generate findings all day. It cannot be the accountable owner of what the organization believes. It cannot look at a two-year-old belief, decide it is now wrong, and stake a roadmap on the correction. That is judgment, and judgment is the entire product.

So the prediction is that UXR’s value relocates almost completely onto the frame. Owning it. Keeping it current. Keeping it honest. The researchers who make it through the next few years are the ones who stop describing themselves as people who run studies and start describing themselves as the people accountable for whether the org’s model of its users still holds.

And here is where it turns operational, and slightly ominous.

A model is only worth the evidence underneath it. The frame has a property I keep calling confidence, which is just the honest answer to “how do we actually know this, and how much of it is folklore somebody said in a meeting in 2023.” The moment real decisions start riding on AI-synthesized evidence, and they already do, that question stops being academic. Somebody has to own the chain. Where did this claim come from. What supports the number the VP just repeated to the board as if it were scripture.

Right now that chain is held together by vibes and one researcher’s memory. That does not survive scale.

So somebody is about to get a job title with the word “provenance” in it. “Evidence governance.” “Research provenance.” Whatever name ends up sticking. Possibly an audited one. The mechanism is boring and predictable. Some company makes a loud, expensive, public decision on synthetic data. It goes sideways in a way that ends up in a deck titled “Learnings.” And suddenly provenance has a budget, a headcount, and a dashboard nobody opens.

Privacy got formalized after its disasters. Data governance got formalized after its disasters. The frame is the asset. Provenance is the audit trail that keeps it from quietly turning into fiction. Both are next in the queue, and the queue is moving.

The study dies. The query lives.

Last big one. It follows from something I will not stop saying, which is that findings have a half-life. They decay. The thing you learned last spring is not a fact. It is a fossil, accurate about a world that has since stood up and walked off.

If that is true, then the whole shape of the work is wrong.

The six-week study. The intake form. The prioritization meeting. The staffing. The delivery. The readout everyone claps for and nobody opens again. That entire machine is built for a world where knowledge sits still.

Knowledge does not sit still.

So the deliverable changes. It stops being the readout, the event, the deck you present once and bury in a folder named “Q3 Final FINAL v4.” It becomes a living thing you maintain and ask questions of. The unit of work stops being the project and becomes the question. You do not commission a study. You ask the system, and the system answers from everything you already know, and it tells you how stale that knowledge is while it does it.

Research ops as we practice it today does not survive that. I am not sure research ops has noticed.

So what is the actual prediction here

Here is where I land.

There are two stories about all of this, and only one of them is true for most of us.

Story one is the nice one. Everything converges, the silos fall, and a unified intelligence function rises to take its seat at the strategy table, respected, central, finally understood.

Story two is the one I keep coming back to. The field splits. Cheap, ambient research everywhere, in every tool, available to anyone who can type a question. And a small, re-credentialed core doing the slow, hard, human work that refuses to automate, the kind that quietly becomes a luxury good. Not one function. Two tiers. And given how organizations behave the second the budget tightens, most of them land on the cheaper tier.

I do not think the future of UXR is doom. I think it is bifurcation. Which is worse, because doom is at least simple. You leave and you learn to bake.

Bifurcation makes you choose which half you are running toward, while the middle stands there insisting there is plenty of time to decide.

There is not.

And none of this needed a crystal ball. That is the part that should bother you. Every prediction here is something the field already half-knows and works hard not to look at. The future of UXR is not hidden. It is just inconvenient.

🔮 If this landed, subscribe to The Voice of User. Pick a tier before the market picks one for you.