How to get your first 100 users for an AI SaaS

Why the first 100 users are a research sample, not a growth number

The first 100 users are a small, hand-picked sample you use to learn whether your product solves a real problem for a real person, not a milestone you hit on the way to a bigger number. Treating them as a growth number leads founders to optimize for the wrong thing at exactly the moment when the wrong optimization is most expensive.

If you chase 100 signups from a viral post or a broad ad campaign, you can hit the number in a day and learn almost nothing, because the people who clicked did not necessarily have the problem you built for, they had five seconds of curiosity. A hundred users acquired this way often produces worse signal than twenty users acquired one conversation at a time, because the twenty were each individually qualified by a human before they ever saw the signup button.

The right question at this stage is not "how do we get to 100 fast" but "who are the 100 people whose feedback would actually change what we build next." That reframes every channel decision that follows. A channel that produces 300 low-intent signups and zero usable feedback is worse than a channel that produces 40 signups where half of them reply to your onboarding email with a specific complaint you can act on.

This is also why the earliest users should skew toward people you can talk to directly. A DM conversation, a five-minute call, or a reply to a personal email gives you ten times the information of a silent signup. Structured, founder-to-founder feedback platforms exist for exactly this reason: getting a stranger who has actually built something to try your product and tell you what confused them is worth more than a thousand impressions from people who never open the app.

Rule of thumbOptimize the first 100 for what you learn from them, not for how fast you reach the number.

Define activation before you send a single message

Activation is the one specific action inside your product that proves a new user experienced the value you promised, and it has to be defined and instrumented before outreach starts, not discovered afterward by looking at a confused analytics dashboard.

For an AI SaaS, activation is rarely "created an account." It is usually something closer to "generated one usable output," "connected one data source and saw a real result," or "completed the first workflow end to end." Pick one moment, write it down as a sentence, and make sure your analytics can answer the question "what percentage of signups reached this" on demand, not by manually digging through logs three weeks later.

Once activation is defined, every channel and every message can be judged by the same yardstick: not how many people signed up, but how many people activated. This single change in what you measure will reshape which channels look promising and which ones you should have abandoned in week one.

• ▸Write activation as one sentence: the specific action, not a vague feeling of engagement.
• ▸Instrument it before outreach, so day one traffic already produces usable data.
• ▸Report signups and activation together everywhere, never signups alone.
• ▸If two different channels produce the same signup count but different activation rates, the channel is not equal even if it looks equal on a spreadsheet.

Founder communities: the highest signal, lowest volume channel

Founder communities are spaces where other builders, indie hackers, and small teams hang out to trade feedback and support each other's launches, and they are the highest signal channel available to a very early AI SaaS because the people in them are used to trying rough products and telling you the truth.

The mistake most founders make in these spaces is showing up once with a link and disappearing. Communities like these run on reciprocity: comment on other people's launches, answer questions in your area of expertise, and be visibly useful for a week or two before you post your own thing. When you do post, frame it as a request for feedback on a specific question, not a generic "check out my product," because specific requests get specific, useful replies.

Expect volume here to be small, often five to twenty real users per community over a few weeks, but the quality of what you learn is disproportionately high. This is also where LaunchLoop fits naturally: a listing gives you a structured place to collect that founder-to-founder feedback and a built-in audience of people who expect to try something new and comment honestly, instead of relying entirely on cold posting into forums where you have no existing reputation.

• ▸Give value in the community for one to two weeks before you post anything of your own.
• ▸Ask a specific question, not a generic ask for upvotes or traffic.
• ▸Reply to every single comment, even critical ones, within hours.
• ▸Expect low volume, high signal: 5 to 20 quality users per community is a good outcome.

Niche communities and forums: go where the problem already gets discussed

Niche communities are subreddits, Discord servers, Slack groups, and forums organized around the specific problem your product solves rather than around startups in general, and they matter because the people in them already talk about the problem, which means your product does not need to be introduced from zero.

Search for the exact phrases people use to describe the problem your AI SaaS solves, not your product category. Someone frustrated with manual invoice reconciliation searches "how do I stop doing this by hand," they do not search "AI accounting SaaS." Find the threads where people are already complaining or asking for a workaround, and answer with genuine help first, mentioning your product only if it is directly relevant and you disclose that you built it.

The rule that keeps this channel from backfiring is simple: never post your product as the first message in a thread you did not start, and never post in a community without reading its self-promotion rules first. A single rule-breaking post can get you banned and burn the channel permanently, which is a bad trade for a handful of clicks.

• ▸Search for problem language, not product category language.
• ▸Answer with real help first, disclose that you built the tool when you mention it.
• ▸Read each community's self-promotion rules before posting anything.
• ▸One well-placed, honest answer in an old thread can bring in users for months through search.

Directories and launch platforms: compressed attention, not a magic number

Directories and launch platforms are places built specifically to surface new products to an audience that opts in to look at new things, and their value at this stage is a short window of concentrated attention rather than a guaranteed number of signups.

Treat every listing as an experiment with a specific hypothesis, not a lottery ticket. Write the listing for the ten people who matter most, not for maximum breadth: a sharp, specific headline that names who the product is for and what it replaces will out-convert a broad, exciting-sounding one every time, because it filters for the right visitors instead of maximizing raw clicks.

LaunchLoop is built around this exact stage of a company's life: it gives an AI SaaS founder a place to launch to an audience of other builders, collect structured feedback rather than silent traffic, and come back for a relaunch a few months later once something real has changed, instead of treating a single listing as a one-shot event. Because the audience is other founders, the feedback tends to be specific and technical, which is exactly what you want when you are trying to learn from your first 100 users rather than just count them.

Spread listings across two or three directories over a few weeks rather than all on the same day, so you can tell which platform's audience actually converts into activated users, and so a bad first attempt on one platform does not use up your only shot.

• ▸Write the headline for the narrowest true description of your buyer, not the broadest.
• ▸Space listings out across platforms so you can compare which audience converts.
• ▸Reply to every comment on the listing within hours, the same discipline as community posts.
• ▸Save a relaunch for a real milestone: a new feature, not a repeat of the same pitch.

Cold outreach: small volume, high personalization

Cold outreach at the first 100 users stage means personally messaging a short list of people you have identified as likely to have the exact problem your product solves, and it works only when the volume stays small enough that every message can be genuinely personalized.

Build a list of 30 to 50 people, not 3,000. Use LinkedIn, Twitter, niche Slack or Discord member lists, or even a public roster from a relevant community, and for each person write one sentence that proves you looked at what they actually do. A message that references someone's specific job, their recent post, or a tool they mentioned using converts at a completely different rate than a templated blast, because it reads as a human noticing them rather than a script finding them.

Keep the message short: who you are in one line, why you are messaging this specific person, and a low-effort ask. Do not ask for a fifteen minute call in the first message, ask a yes or no question they can answer in five seconds, like whether the problem you describe is one they actually deal with. That small first ask has a far higher reply rate and gives you a foothold to follow up with the product.

• ▸List size: 30 to 50 people, hand-picked, not a scraped mass list.
• ▸One personalized sentence per message, referencing something specific and true about that person.
• ▸First ask should be a five-second yes-or-no question, not a scheduling request.
• ▸Send in small batches of 10 to 15 so you can adjust the message based on early replies.

A cold outreach script that gets replies

A cold outreach script is a short, reusable message template with two or three variables you fill in per person, and having one ready removes the friction that keeps most founders from sending messages at all.

Here is a template that works because it is short, specific, and asks for almost nothing: "Hi [name], I noticed you [specific detail: posted about X, work at Y, mentioned Z]. I am building [one line description] because [specific problem]. Is this something you deal with, or am I off base?" That is four sentences, it takes under a minute to personalize, and the question at the end is easy to answer even from a phone.

For people who reply yes, the follow up should invite them in without pressure: "Would you be up for trying an early version and telling me honestly what is broken? No pitch, I just want it to actually be useful." Framing it as a request for honesty rather than a sales pitch changes how people respond, because you are asking them to help rather than asking them to buy.

For people who do not reply within four or five days, one polite follow up is reasonable, a second one is usually not. Move on and spend the time on new names instead of chasing the same ones repeatedly.

Rule of thumbThe best cold message asks for an opinion, not a signup. The signup comes after they say yes.

Content and search: slow to start, compounding once it works

Content for early user acquisition means writing pages, in your own voice, that answer the exact question your future users are typing into search engines and AI assistants, and it is the slowest channel to produce its first user but the only one that keeps producing users without repeated manual effort.

At the first 100 users stage, do not write generic thought leadership. Write pages that answer a narrow, specific question your ideal user has right now, such as a comparison, a how-to, or a troubleshooting guide for the exact workflow your product replaces. This is also increasingly how AI answer engines find sources to cite, so a page written to genuinely answer one question well tends to get pulled into both search results and AI-generated answers over time.

Expect content to contribute a small number of your first 100, often in the single digits in the first month, because search and AI citation both take time to index and trust a new domain. Its real value at this stage is compounding: the five pages you write now become the twenty pages that bring in your next 500 users without any additional outreach effort.

• ▸Write for one specific question per page, not a broad topic.
• ▸Prioritize pages that match what your ideal user actually searches, not what sounds impressive.
• ▸Treat early content traffic as a bonus, not a plan you depend on for the first 100.
• ▸Reuse the language from your cold outreach and community replies, since it is proven to resonate.

Partnerships and integrations: borrow someone else's trust

A partnership at this stage is an agreement with another small tool, newsletter, or community that already has the audience you want, where you offer something of value in exchange for a mention, an integration, or a joint post, and it works because it borrows trust that would otherwise take months to build from scratch.

Look for tools that are adjacent to yours, not competitive, meaning your ideal user probably already uses that tool for a related task. A simple integration, even a basic one built in a weekend, can be enough to justify a mention in their changelog or newsletter, which reaches an audience that is already primed for exactly the kind of product you are offering.

Newsletter writers and small community moderators in your niche are also worth approaching directly with a straightforward offer: early access in exchange for their honest opinion, shared with their audience only if they actually like it. Do not ask for a guaranteed positive mention, ask for an honest one, because forced positivity is easy for readers to detect and it damages both your and their credibility.

• ▸Target adjacent, non-competing tools your ideal user already relies on.
• ▸A weekend integration can be worth more than a week of cold outreach if it lands the right partner.
• ▸Offer early access for an honest opinion, not a guaranteed positive post.
• ▸Track which partner's audience actually activates, not just which partner agreed to mention you.

Designing onboarding so hard-won users actually activate

Onboarding is the sequence of steps a new user goes through between signing up and reaching your defined activation moment, and at the first 100 users stage it deserves as much attention as acquisition, because a user you fought hard to bring in who drops off during setup is a wasted effort twice over.

Count the number of steps between signup and the first real output, and cut anything that does not directly serve reaching that output faster. Every additional field, every optional setting shown before the user has seen value, is a place where an early, unforgiving user quietly closes the tab and never comes back. AI products in particular often ask for too much configuration up front when a single sensible default would let the user see a result in under a minute.

Because your first 100 users arrived through personal channels, you have an option most companies at scale do not: manually checking in on each one during their first session. A short message sent an hour after signup, "did you get to try it yet, happy to help if anything is confusing," catches people at the exact moment they are stuck and about to give up, and it turns a silent drop-off into a saved user and a piece of feedback.

• ▸Count and cut onboarding steps that do not lead directly to the first real output.
• ▸Prefer sensible defaults over configuration screens for the first session.
• ▸Manually check in on early signups within the first hour or two while volume is still small.
• ▸Watch for the exact step where users drop off, that step is your highest priority fix.

Measuring retention signals that predict a real product

A retention signal is any piece of evidence that a user came back and used the product again without being prompted, and among the first 100 users these signals matter more than the total count, because they are the earliest honest evidence that you have built something worth scaling.

Watch week 2 retention specifically: the share of activated users who return and use the product again seven to fourteen days after their first session, on their own, without a reminder email from you. This number is small and noisy at this sample size, so read it directionally rather than as a precise statistic, but a pattern of near-zero unprompted returns across your first 30 or 40 users is a real warning sign worth stopping to investigate before you push harder on acquisition.

Also pay attention to qualitative retention signals that do not show up in a dashboard: a user asking when a specific feature is coming, a user inviting a colleague without being asked, a user replying to your check-in with a paragraph instead of a one-word thanks. These are small but they are honest, unprompted signs of real interest that a raw signup count cannot tell you.

• ▸Week 2 retention: who came back on their own, without a nudge, is the core number.
• ▸Read the number directionally at this sample size, not as statistically precise.
• ▸Watch for unprompted, qualitative signals: questions about future features, unsolicited referrals.
• ▸A pattern of weak retention across many early users points to a product gap, not a distribution gap.

When to stop a channel instead of grinding on it

Stopping a channel means deliberately ending a specific acquisition effort because it has produced enough evidence to judge it, rather than abandoning it out of frustration or continuing it out of stubbornness, and having a clear rule for this decision saves the weeks that founders otherwise lose to channels that were never going to work.

Set the rule before you start, not after you are three weeks in and emotionally invested. A reasonable default is two weeks and a defined volume, for example 40 cold messages sent or two community posts made, with a clear bar for what counts as working: a minimum reply rate, a minimum number of activated users, or some other number you write down in advance. If the channel does not clear that bar in that window, stop and move the time to a different channel rather than doubling the volume on the same approach and hoping the outcome changes.

The exception to stopping quickly is a channel that is producing strong qualitative signal even at low volume, such as founder communities where five deeply engaged users are worth more than fifty shallow ones from elsewhere. Judge channels by the quality of what they produce against your activation and retention numbers, not by raw signup count alone, and you will make this call correctly far more often.

Rule of thumbDecide your stop rule before you start the channel, so a bad week does not get mistaken for a bad channel, or a bad channel does not get mistaken for a bad week.

Bringing it together: a realistic four week plan

A realistic four week plan for the first 100 users spreads effort across two or three channels at once rather than betting everything on one, because different channels reveal different kinds of problems and no single channel is likely to carry the whole number on its own.

Week one: instrument activation, prepare your outreach list and message template, and start giving value in one or two founder communities before posting anything. Week two: send your first batch of cold outreach, post in the communities where you have built some presence, and prepare a directory listing with a specific headline. Week three: launch the directory listing, follow up personally with everyone who signed up in the first 48 hours, and start writing your first one or two content pages based on the language you have heard repeatedly in outreach replies. Week four: run your retro, look at activation and week 2 retention across every channel side by side, decide which channels earned another round and which get cut, and plan a relaunch or a second push only for the channels that cleared their bar.

By the end of four weeks you should have a small, honest picture: which channel produced users who actually activated and came back, what language and framing got replies, and a concrete list of onboarding fixes straight from real sessions. That picture, not the raw count of 100, is what actually tells you whether you have a product worth scaling.