What AI Taught Me About Markets, Users, and Myself

Discovery is a conversation.

Over the past few months, I’ve found myself turning to AI not just for writing help or brainstorming—but for something bigger: discovery. Of markets. Of users. Of myself.

Here are three stories of how AI helped me think better, ask better questions, and ultimately make better decisions.

First, I used AI to study the market.

I’ve been curious why it feels so hard—logistically and emotionally—for women to get outside, whether it’s backpacking, hiking, surfing, skiing, or something else. I kept hearing safety come up, but I didn’t know how widespread that concern really was—or what was underneath it.

So I turned to OpenAI’s Deep Research. I gave it my hypothesis, outlined my target demographic, and asked it to dig in.

The opening paragraph was fantastic:

I think about the highlighted stat from Outside Magazine all the time. Not because I was surprised that women were mostly afraid of men—but because it confirmed something I’d felt and heard, and had never quite articulated. It’s now the stat I lead with when I explain this space to others.

It also pulled interviews from a range of women—experts and everyday people. Some of it was anecdotal, but it was incredibly validating. I think sometimes the most useful part of research isn’t the data: it’s the resonance of feeling like you’re on the right track. See the snippet below:

Then it gave me something actionable: a competitive scan of existing solutions. Here’s a snippet on personal safety devices and apps:

I’ve found Deep Research is a substitute for a sharp junior analyst who could map out the “market landscape & competitive” section of your PRD. It won’t catch every edge case, and you’ll still need to validate with real users—but as a starting point, it’s absurdly useful.

Below are some of my tips from using Deep Research:

• Don’t overthink the prompt. OpenAI will usually ask clarifying questions before it runs the search, and you can always tell it, “Help me write this prompt better.”

Sample follow up questions from OpenAI before Deep Research runs

• Ask fewer questions. I found 4–5 questions max yields the best results. When I got greedy, the output was scattered—interesting, but shallow.

• Don’t hoard your queries. I was being very precious with my credits. But they refresh and increase often, and the more you ask, the better you understand the edges of the tool.

Then, I used AI to do user research.

At a product event, I watched a CEO demo an AI tool that ran full user interviews—with real people. It was better than I expected. The users were so engaged, and able to speak in their native languages, in time zones and places that worked best for them. It was scalable in a way no human research team could match.

I wasn’t quite ready to send my own users—mostly friends—into a solo interview with Claude or GPT. But I figured AI could at least help me write better user interview questions.

And when it crushed that task, a new idea hit me: what if I used it to interview myself?

So I gave it everything: my target audience, my hypothesis, my assumptions, a few discovery frameworks I like. Then I told it to play the role of researcher.

At first, it returned a wall of questions. But the real unlock was asking it to go one question at a time.

I love that it said, “You don’t have to perform, just answer naturally.”

As it asked questions, something clicked. I said something I’d never verbalized before: with backcountry skiing, the thing I’d actually pay for wasn’t better gear, maps, weather data, or partner finding—it was a human validator for routes. Someone to say, “Yes, that tour looks safe.” See the snippet below:

I love how I just answered totally different questions sometimes.

It asked me about every activity I do—from solo backpacking to all kinds of skiing to trail running. When we wrapped, I asked it to summarize what it had learned. The summary was useful—but the real value was in the questions. Being interviewed forced me to slow down, explain my thinking clearly, and see my own patterns more objectively. It helped me understand myself as a user with needs.

The biggest surprise was that it made me better at interviewing. It demonstrated what a great user interview looks like, and even though I knew many of the practices, it’s much easier to get better by copying and getting suggestions on the actual user interview you’re doing, not hypothetical situations.

Then, I used AI for therapy.

This was one of those moments that truly caught me off guard.

I journal regularly and one of the styles of journaling I do is asking myself questions. So after I understood how good it was at user discovery, the natural extension was to use it for therapy.

I fed it a few of my recent journal entries and asked it to help me explore what am I actually looking for from dating.

This is a snippet from early in the conversation:

This framing caught me completely off guard. I hadn’t intentionally prompted it to ask about presence vs. possibility, nor had I even thought about that—but it was definitely the core tension I was feeling.

The conversation continued and I was made to realize how many deep sources of love I had in my life.

Later, when I asked it how I could hold on to that clarity, it gave me this:

Sure, I could have had this conversation with a friend or hired a therapist. But what made this moment powerful was that in that exact moment, I could chat and not worry about how I was coming across.

That conversation didn’t just soothe me—it shifted me. It’s wild to realize we’re at a point where AI doesn’t just make life faster or more efficient… it can actually help you think differently. I think about Isaac Asimov and Philip K. Dick who formulated so much of my viewpoint on AI, and I wish they were here to write me more stories to figure out how to take in all of this.

It’s breathtaking.