10 Signals Your AI Product Is Approaching PMF

PMF Is a Pattern, Not a Moment

Founders often describe achieving product-market fit as a sudden revelation — the moment everything clicked. In reality, PMF accumulates gradually from dozens of small signals before it becomes undeniable.

The problem with waiting for the revelation is that you might miss the signals that precede it. Or worse, you might mistake one good signal for the full picture and stop iterating too early.

We watch 10 signals across our portfolio companies when evaluating PMF progress. No single signal is definitive. The pattern across all of them tells the story.

Signal 1: Retention Curves Flatten

The most reliable technical signal of PMF is a retention curve that flattens rather than falling to zero.

Every product loses users over time. The question is whether it loses all of them or whether a subset sticks.

If your 30-day retention curve continues to fall toward zero, you don't have PMF — at least not yet. If it flattens at even 20-30%, that means some users find permanent value. Find out who those users are, what they're using the product for, and build for more of them.

Signal 2: Users Are Upset When the Product Is Down

This one sounds obvious, but it's not. Most products are not missed when they go down. Users find another way to do the thing.

When your users notice downtime immediately, contact you within minutes, and express genuine frustration — that's a signal the product is integrated into a workflow they depend on. Dependency is a form of value.

Track your inbound contacts during outages. A product approaching PMF generates disproportionate outage feedback.

Signal 3: Unsolicited Referrals

You're not asking people to refer others. They're doing it anyway. A user tells a colleague, "You have to try this." Another shares a screenshot in a community they're part of.

Unsolicited referrals are the highest-quality signal of PMF because they represent revealed preference. The user is spending social capital — their reputation — to recommend your product. People only do that when they genuinely believe in what they're recommending.

Track where your signups come from. When you start seeing a meaningful percentage with "referral from a friend/colleague" or similar as their source, pay attention.

Signal 4: Users Build Workflows Around Your Product

Watch how users use your product. Are they integrating it into their daily or weekly workflow? Checking it first thing in the morning? Building other tools or processes around it?

When a product becomes a dependency — not just a nice-to-have but a load-bearing part of how someone works — that's a strong PMF signal.

For AI products specifically: are users saving their prompts? Building custom integrations? Getting angry when the AI outputs change? Workflow integration is extremely positive.

Signal 5: Paying Customers at a Price That Covers Your Costs

Willingness to pay is one of the sharpest signals available. Talk is cheap; credit cards are not.

But not just any paying customers. The question is whether users are paying at a price that makes the unit economics work. Free users who convert to paid at $1/month and immediately churn are not a PMF signal. Customers who pay $50/month and stick for six months are.

The pricing itself matters. If you can't charge what you need to charge to make the business work, you don't have PMF for this customer segment — even if they're enthusiastic users.

Signal 6: Session Depth Is Growing

Users aren't just logging in; they're doing more each time they visit. Session depth (pages visited, features used, time spent on meaningful actions — not just idle time) grows as the product becomes more central to their work.

For AI products, session depth might look like: number of conversations per session increasing, number of documents processed per session increasing, or time from session start to first meaningful output decreasing (they know exactly what they want).

Signal 7: A Specific Use Case Is Crystallizing

Early in a product's life, different users use it for different things. PMF often comes when you notice a cluster of users using it for one specific thing — and loving it.

This crystallization is important information. It tells you who the product is really for and what problem it's actually solving — which is often different from what you thought when you built it.

The temptation is to keep building for all the use cases. The PMF-seeking move is to double down on the crystallized one, even if it means de-emphasizing others.

Signal 8: Organic Word-of-Mouth Without Marketing

Your paid acquisition is turned off (or you haven't turned it on yet). Your content calendar is empty. Your growth is still positive.

Organic growth means the product itself is generating awareness. Users are talking about it, sharing it, writing about it. This is the highest-quality growth signal because it's zero-cost and self-sustaining.

If you're relying exclusively on paid acquisition to grow, you might have a marketing problem you're mistaking for product-market fit.

Signal 9: Inbound from Your Target Segment

You start getting emails and signups from people who match your ideal customer profile exactly — but they found you without any outbound from you. Maybe they heard about you from a conference, a community, a blog post by a user.

Inbound from the right people is qualitatively different from viral growth among the wrong people. Both are growth, but only one of them means you've found PMF for a segment that matters.

Signal 10: Team Energy

This one is subjective, but real. When a product is working, the team knows it. Customer calls are exciting rather than deflating. Support tickets reveal interesting edge cases rather than fundamental misunderstandings. The team has a clear intuition about what to build next because they hear consistent feedback from enthusiastic users.

When a product isn't working, the team often knows that too — there's a background anxiety, difficulty aligning on priorities, support tickets that feel like attacks rather than feedback.

Team energy isn't a substitute for data. But experienced founders who trust their gut are often picking up on real signals in aggregate that haven't yet shown up clearly in metrics.

How to Use These Signals

No product will show all 10 signals simultaneously before it "counts" as PMF. But if you have 0-2, you probably don't have it. If you have 7-8 strongly, you probably do. The 3-6 range is where the real work happens — figuring out which signals are weak and why.

Use these signals as a diagnostic, not a checklist. They point to where the product is working and where it isn't, which is exactly the information you need to iterate toward fit.