Opinions expressed by Entrepreneur contributors are their own.
Key Takeaways
- AI speeds up research tasks, but human judgment still drives insight.
- Use AI for sorting data, not deciding what matters.
- Faster research means insights arrive before product decisions are finalized.
For most of my career, the research process I relied on had one built-in assumption: that you’d have time. Time to recruit, run sessions, sit with transcripts for a few days and let the themes show themselves. I spent eight years doing that across telecommunications, financial services and enterprise software, and it worked.
Fast-paced, sure, but predictable enough that a well-scoped study could keep up.
That changed when I started leading research for AI-powered calling and collaboration products used by more than 80 million people. The race toward AI fluency hit hard in 2025, and suddenly, the product teams I support were moving faster than my timelines could handle.
I had a choice: rethink how I work, or keep delivering insights after the decisions were already made. So I started experimenting with AI in my own workflow. What I’ve found is that it can genuinely speed up the mechanical side of research without watering down the rigor. But you have to be intentional about where you bring it in and where you don’t.
Planning: lay stronger foundations in less time
Every study starts with groundwork: desk research, literature reviews, drafting a plan, figuring out what stakeholders need answered. This phase used to eat up days before I could begin the real investigation.
Now, I use Claude to pull together structured lit reviews, surfacing academic work, industry reports and prior findings much faster than manual searching. I still do the hard part: layering in internal context, deciding what’s relevant and identifying gaps that primary research needs to fill.
For research plans, I have AI generate a first draft, then sharpen it with specifics that only come from knowing the product space.
Copilot has been a game-changer here, too. When research comes up during cross-functional meetings, I use it to capture and summarize discussions so nothing falls through the cracks. When you’re supporting nine teams and more than 30 stakeholders, that kind of real-time capture matters more than people realize.
Execution and analysis: let AI handle the volume so you can focus on meaning
This is where I’ve seen the biggest time savings, and where you need to be the most careful. Qualitative synthesis used to be my biggest bottleneck. I’d spend days coding transcripts, mapping themes and pulling out quotes.
Necessary work, but a lot of it is sorting rather than thinking. I now use Marvin AI to generate initial thematic maps and do a first pass at coding. What used to take days shows up in minutes.
On the quant side, Copilot’s Analyst Agent handles large survey datasets, sometimes thousands of responses, pulling out directional themes way faster than I could alone.
But here’s what really matters: what happens after the AI gives you its output. It can surface patterns all day long. I’m the one who decides which patterns actually matter. I know what the product team is wrestling with, what the business strategy calls for and where our assumptions might be wrong.
The sorting got faster. The thinking? That’s still on me.
Shareout: make insights land when decisions are still open
Research is only worth something if it reaches the right people while they can still act on it. Once I cut down my synthesis timelines, stakeholders began pulling me into conversations earlier. Not because my work was better, but because it was ready when things were still being figured out. I lean on Copilot across M365 to get draft structures, slide layouts and summary narratives together quickly, then add the storytelling that makes people actually pay attention.
For weekly executive leadership updates, AI lets me share directional data while it’s still useful instead of waiting for a polished final report.
That, to me, is the strongest argument for AI in research. It won’t make you a better researcher. But it’ll make you a faster one. And when your teams ship every few weeks, speed is what separates research that shapes the product from research that just documents what already went out the door.
Beyond the research: extend your reach as a researcher
There’s a whole layer of work beyond user research: the planning, logistics and visibility work that eats into your week.
I use AI to plan my weeks across teams and build visual roadmaps so partners see where my time is going. When you’re a shared resource, keeping people informed is how you keep trust intact.
Looking further out, I think the next big opportunity is building AI agents for your workflows. I’ve been exploring this with Claude Code: designing agents that could parse evaluation data, keep deliverables searchable, cross-reference old studies so teams don’t duplicate work and help product partners figure out when to engage research.
Every one of these would let a single researcher have a wider impact without being in every room at once.
The line that matters
I don’t think the researchers who come out ahead will be the ones who adopt AI fastest. It’ll be the ones who get clear about where it belongs and where it doesn’t.
AI is great at organizing, summarizing, drafting and spotting patterns.
It cannot tell you what those patterns mean for a specific product at a specific moment. It can’t build the stakeholder relationships that turn a finding into a decision. And it can’t sit with ambiguity the way a researcher can when the data points somewhere the team doesn’t want to go.
That work, the judgment calls, the empathy, the ability to tell a story that moves people to act, that’s the whole point of being a UX researcher. AI just gives you more room to do it well. But only if you hold the line on the parts that need a human mind behind them.
Key Takeaways
- AI speeds up research tasks, but human judgment still drives insight.
- Use AI for sorting data, not deciding what matters.
- Faster research means insights arrive before product decisions are finalized.
For most of my career, the research process I relied on had one built-in assumption: that you’d have time. Time to recruit, run sessions, sit with transcripts for a few days and let the themes show themselves. I spent eight years doing that across telecommunications, financial services and enterprise software, and it worked.
Fast-paced, sure, but predictable enough that a well-scoped study could keep up.
