AI Made Me Talk to 10x More Users. That's the Real Discovery Shift.

I thought AI would help me skip user interviews. That was the pitch, right? Feed it data, get insights, build faster.

Turns out, AI did the opposite. It made me talk to way more users than I ever did before. And that completely changed how I think about discovery.

The old way was slow, so we just did less of it

Before AI tools became actually useful (like, 18 months ago), here's what product discovery looked like for me:

I'd schedule 5-6 user calls a week. Each one took 45 minutes. Then I'd spend another 30 minutes writing notes, tagging themes, trying to remember what mattered. By Friday, I had maybe 15 pages of Google Docs and a vague sense that "people want better reporting."

At Finvestfx, I was managing 20+ enterprise clients in forex and treasury. I knew I should be talking to more of them. But between demos, support escalations, and internal meetings, I just couldn't. So I talked to the loudest clients and assumed they spoke for everyone. Spoiler: they didn't.

The real problem wasn't that discovery was hard. It was that synthesizing it was hard. So I did less discovery to avoid drowning in notes.

AI didn't replace interviews, it made them 10x more useful

At Sonic Linker, we were moving fast. Founding team, AI SaaS product, trying to ship the core platform in 3 months. I couldn't afford to miss signals.

I started using AI tools (mostly ChatGPT and later some transcript analyzers) to do the boring parts. Record a call, dump the transcript, ask it to pull out pain points, feature requests, objections, anything that contradicted what I thought I knew.

Suddenly, I could talk to 15 users a week instead of 5. Because I wasn't spending hours in post-call cleanup. The AI did the first pass. I'd review it, add context, move on.

But here's the part that actually mattered: I started noticing patterns I used to miss.

When you're manually tagging themes across 6 calls, you remember the loud stuff. The thing three people said the exact same way. But when you're analyzing 15-20 calls with AI help, you start seeing the quiet stuff. The offhand comment that five people made in slightly different words. The feature they didn't ask for but kept working around.

At Sonic Linker, we almost built a dashboard feature because two enterprise leads asked for it directly. But when I analyzed a broader set of calls, I realized the real ask was "we need to explain this to our legal team." The dashboard would've been nice. A one-click compliance report was what actually closed deals.

The new discovery loop is talk more, synthesize faster, test quicker

I used to think discovery was about depth. Spend an hour with a user, really dig in, understand their world.

Now I think it's about breadth first, then depth. Talk to 20 users in a week. Use AI to spot the patterns. Then go deep on the ones that matter.

Here's my actual loop now:

1. Talk to everyone. Not just power users. Not just the clients who pay the most. Everyone. I aim for 12-15 conversations a week, even if some are just 20 minutes.

2. Let AI do the first synthesis. I dump all transcripts into a tool (or honestly, just a ChatGPT thread) and ask it to find overlaps, contradictions, edge cases. It's not perfect, but it's fast.

3. I add the context AI misses. Tone, urgency, who they are in the org, whether they're a churner or a champion. AI can't read a room. I can.

4. Test fast. Once I see a pattern across 8-10 calls, I don't write a full spec. I mock it up, show it to 5 more people, see if they light up or shrug.

At Finvestfx, I wish I'd had this loop. I would've caught the retention issues earlier. I was talking to happy clients and assuming the quiet ones were fine. They weren't. They just didn't complain before they churned.

What actually changed

AI didn't replace product discovery. It removed my excuse for not doing enough of it.

I used to avoid calls because the aftermath was overwhelming. Now the aftermath takes 10 minutes. So I take more calls. And the more calls I take, the faster I spot what actually matters versus what's just noise.

If you're a PM and you're not using AI to scale your user conversations, you're leaving insights on the table. Not because AI is smarter than you. But because it makes the boring parts fast enough that you can actually talk to everyone instead of just the squeaky wheels.

The real shift: Discovery isn't a phase anymore. It's just part of the weekly rhythm. And AI is the only reason I can keep up.