okaihub

How to Get Your First 100 Users for an AI Product

A cold-start sequence for a new AI product with no audience: who to ask first, which communities tolerate a launch, and what to do in the 30 days after.

Published 2026-11-02 · 4 min read

The first 100 users of an AI product almost never come from a launch. They come from twenty small, unglamorous actions taken in a specific order, most of which involve talking to one person at a time. This is that order.

Users 1 to 10: do things that do not scale

Your first ten users should not come from a platform. They should come from people you can message directly, because at this stage you need conversations, not traffic. A signup tells you almost nothing; a reply tells you everything.

  1. 1

    Write down 25 people who have the problem

    Former colleagues, people in communities you already read, founders whose products solve an adjacent problem. Not people who might like it — people who have the problem today.

  2. 2

    Message ten of them individually

    One or two sentences. Name the specific thing they posted or built that made you think of them. Ask whether the problem is real for them, not whether they will try your product.

  3. 3

    Set them up yourself

    Do not send a link and hope. Create the account, import their data, run the first task with them on a call. This is slow and that is the point: you will learn more in five onboarding calls than from five hundred signups.

  4. 4

    Ask one question a week later

    "Did you use it this week, and if not, what got in the way?" The honest answers here are the product roadmap.

Users 10 to 40: one community, done properly

Pick a single community where your buyers already gather and become useful in it for two weeks before mentioning your product. Every community has norms, and the fastest way to get removed is to arrive posting a link.

CommunityWhat it toleratesHow to enter
Reddit (a specific subreddit)Detailed first-hand experience. Not links.Answer questions in your niche for two weeks. Mention your product only when someone asks, then explain the trade-offs honestly.
Indie HackersProgress posts with real numbers, including bad ones.Share what you learned and what failed. Self-promotion is accepted when the post is useful without it.
Hacker NewsNovel engineering or original data. Almost no marketing.Show HN for a technically interesting piece, or post a genuine finding. A marketing post will be flagged.
X / TwitterBuilding in public, demos, opinions with evidence.Reply substantively in your niche daily for a month. Your own posts get read once people recognise the handle.
Discord / Slack groupsDepends entirely on the group; many have a showcase channel.Read the rules first. Join and participate before you post anything about your product.
Tolerance varies per group and changes over time — read the current rules before posting anywhere.

Users 40 to 100: directories and listings

This is where listings earn their keep. A directory entry is not a spike — it is a page that keeps describing your product to people who are actively searching for a tool like yours, months after you have stopped promoting.

  • Prepare once, submit many. A logo on a transparent background, a 60-character one-liner, a 160-character description, three screenshots and a short demo video cover almost every submission form. Writing these fresh per platform is how a two-hour job becomes two days.
  • Sort by relevance, not size. A small directory read by your exact buyers beats a large one read by everyone. Check whether the directory's other listings look like your competitors or like noise.
  • Space submissions out. Five directories in one afternoon produces five near-identical announcements and no compounding. One or two a week gives each a chance to be seen, and lets you fix the description based on what the first ones taught you.
  • Answer every comment. The replies under a listing are read by far more people than the listing itself, and an unanswered question is a lost signup.

The full channel-by-channel breakdown, including what each type of listing costs and what it asks for, is in the companion guide on where to list an AI product.

What to do in the 30 days after

Most founders treat the last submission as the finish line. It is the point where the work changes shape: from finding people to keeping the ones who came.

  1. Week 1 — instrument. Know where each visitor came from, and which of them did the core action once. Referrer-based tracking is the fallback for platforms that strip UTM parameters.
  2. Week 2 — talk to everyone who signed up. All of them, while there are only a few dozen. Ask what they were doing the moment before they found you.
  3. Week 3 — fix the drop-off, not the funnel. The step where most people leave is worth more than any new channel.
  4. Week 4 — write down what worked and repeat it once. One channel that produced ten users beats four that produced two each.

What usually goes wrong

  • Skipping the first ten. Going straight to platforms before doing manual onboarding means you optimise distribution for a product you do not yet understand.
  • Too broad a niche. "People who use AI" is not a niche. If you cannot name the subreddit, you cannot find the users.
  • One-day effort. A launch that happens once and is never mentioned again produces a spike and nothing after it.
  • No follow-up. Users who try once and leave are not lost; they are usually one fix away from staying, and you only learn the fix by asking.

More in Cold start