Your First 10 Users Should Not Come From a Launch
Why the first ten users of an AI product should be recruited one at a time and onboarded by hand, and what that week teaches you that a launch cannot.
Published 2026-09-20 · 3 min read
The first ten users are not a distribution problem. They are a research problem wearing a distribution costume, and treating them as traffic is the most expensive mistake available at this stage.
A launch gives you numbers. Ten hand-recruited users give you reasons. At a stage where you still do not know what the product is for, reasons are worth more than numbers by a wide margin.
What a launch cannot tell you
- Why someone left. Analytics shows the step where they dropped; a conversation shows what they were trying to do and what confused them.
- What they called the problem before they found you. Their words are better than yours, and you only hear them if you talk to them.
- What they used instead. The real competitor is usually a spreadsheet, a manual process or doing nothing — not another AI product.
- Whether they would pay, and roughly what for. Not a pricing survey; a conversation where money comes up naturally.
None of this arrives through a signup form. It arrives in thirty minutes on a call, or in a thread of messages where someone describes their actual workflow.
How to recruit ten people properly
- 1
Find people who already have the problem
Look for the workaround, not the keyword. People asking how to do a task manually, people complaining about an existing tool, people whose job listing describes the work your product automates. Write down where you found each one — that is a map of where more of them live.
- 2
Contact them individually with a specific reason
Two sentences: the thing they posted or built that made you think of them, and a question about how they currently handle it. No link, no request to try anything. This gets replies because it is a conversation, not an ask.
- 3
Ask about their workflow before mentioning the product
How do you do this today? How long does it take? What is the worst part? What have you tried? Twenty minutes of this is worth more than any survey.
- 4
Then offer to set it up with them
Not "here is a trial link" — "can I set this up with you on a call and watch you use it?" You are asking for their time, and you are getting observation in return.
- 5
Follow up a week later with one question
"Did you use it this week, and if not, what got in the way?" The honest answers are your roadmap for the next month.
What to do with what you learn
| What you collect | Where it goes |
|---|---|
| The exact phrases people use for the problem | Your one-liner, your description, your tags, your landing page headline |
| Where each person hesitated or asked a question | Onboarding changes, in order of how many people hit them |
| What they used before | Your comparison page and your positioning against the real alternative |
| Who they said else has this problem | Your next twenty prospects |
| Which part they used repeatedly | The core feature — probably not the one you thought |
Why this scales better than launching first
Ten users recruited this way produce the raw material for a launch that works: a description written in buyers' language, onboarding that does not lose people, one or two quotable results, and a list of people who will genuinely look at the product on the day.
Launching first produces the opposite. You spend your one good first impression on a page written in your own vocabulary, pointing at an onboarding flow you have never watched anyone use, in front of an audience you cannot follow up with because you do not know who they are.
When to move on
Ten is enough when you can answer three questions without guessing: what do people call this problem, what do they do today, and what is the first thing they should see when they arrive. If you can answer all three in other people's words rather than your own, you are ready to stop recruiting one at a time and start distributing.
The goal of the first ten users is not ten users. It is knowing what to say to the next hundred.
More in Cold start
- The 10 Mistakes That Kill an AI Product LaunchA diagnostic list of the ten failures behind a flat AI product launch, ordered by how often they happen and how cheaply each one is fixed.
- What to Do in the 30 Days After Your AI Product LaunchThe month after a launch is when most founders stop and most of the value is created. A week-by-week plan for converting a spike into something durable.
- How to Cold Start an AI Tool in a Crowded CategoryWhen thousands of products claim the same category, generic positioning is unrecoverable. How to narrow until the competition stops being direct.