The 90-Day AI Product Growth Plan
A sequenced 90-day growth plan for a new AI product, tying together cold start, distribution, content and measurement into one ordered path.
Published 2026-08-24 · 4 min read
Growth advice arrives as disconnected tactics — post here, list there, write this — with no sequence. The sequence is most of the value, because the same tactic produces completely different results depending on what came before it. This is a ninety-day plan that puts the tactics in the order that works.
It assumes a working product and roughly twenty hours a week of founder time. Adjust the pace, not the order — the dependencies between phases are the point.
The shape of the ninety days
| Phase | Weeks | Goal | Primary metric |
|---|---|---|---|
| Learn | 1-3 | Talk to users; find the language and the leak | Conversations held |
| Establish | 4-7 | Build the durable surfaces; get listed | Activation rate |
| Amplify | 8-11 | One launch or push, into an audience that exists | Right-audience traffic |
| Compound | 12-13 | Start the loop that runs without pushing | Whether the floor rose |
Phase one: learn (weeks 1-3)
The temptation is to start distributing immediately. Do not. Everything downstream depends on knowing who the buyer is and what they call the problem, and three weeks of conversations produces that better than any amount of guessing.
- Talk to twenty people who have the problem. Individually, about their workflow, not your product. Record their exact words.
- Onboard three to five by hand. Watch them use it. The place they hesitate is your activation problem, and finding it now is worth more than any traffic.
- Define the activation event. The one action that separates users who stay from those who leave. Everything in phase two optimises for this.
- Write nothing public yet. The output of this phase is the language and the diagnosis you will use everywhere else.
Phase two: establish (weeks 4-7)
Build the durable surfaces before you drive traffic to them. A spike that lands on nothing is wasted; a spike that lands on listings, a clear page and a comparison converts.
- 1
Fix the activation leak
The single worst drop-off from phase one. One change, measured. This matters more than anything else in the ninety days.
- 2
Rewrite the description in user language
The headline, the one-liner, the first sentence — all using the words from phase one rather than your category vocabulary.
- 3
Build the asset pack and get listed
Ten to twenty relevant directories, spaced out, link type verified. These are your durable, indexable surfaces and the seed of the listing loop.
- 4
Publish the first three pages
The specific-job page, one honest comparison, and the product page done properly. Server-rendered, canonical, crawlable, AI-crawler-permitted.
- 5
Start participating in one buyer community
Answers only, no product mentions. The goal is a recognisable handle by phase three.
Phase three: amplify (weeks 8-11)
Now — and only now — drive concentrated attention, because there is finally something for it to land on and a description that has been tested.
- One launch-day platform, done properly, if the product is demo-able and you can be present all day. You get one good first impression; spend it now, not in week one.
- Post the launch thread built from the phase-one language, with the limitation named and the link near the end.
- One substantive community post, written for that community, mentioning the product in passing.
- Pitch two newsletters and one journalist, with the original angle or data you have. Not a blast — two targeted pitches.
- Answer everything, all day, every day during the push. The follow-through is where most of the value is, and where most founders disappear.
Phase four: compound (weeks 12-13)
The push decays. What remains is whether you built anything that runs without you. The last phase starts the loop.
- Follow up with everyone who engaged in phase three and did not convert. One personal line each. This is the warmest list you will ever have and it is almost always ignored.
- Publish the launch story — what you did, the real numbers, what you would change. It is content, it earns links, and it is honest.
- Start the content loop: one page every two weeks, each answering a question you were actually asked in phases one and three.
- Measure whether the floor rose. Compare your weekly baseline now to week one, with the spikes removed. A higher floor is the real result of the ninety days.
- Decide the next ninety days from what the data says worked, not from what felt busy.
What each phase produces
| Phase | Lasting output |
|---|---|
| Learn | The buyer's language, the activation diagnosis, your first design partners |
| Establish | Fixed activation, live listings, three ranking pages, a recognised handle |
| Amplify | A permanent product page, reviews, a warm list, a story |
| Compound | A content loop, a rising baseline, and the data to plan the next quarter |
The three rules underneath the plan
- Durable before spike. Never drive concentrated attention to a surface that will not still be there next month. Establish before you amplify.
- One channel at a time gets your full attention. The plan sequences channels rather than running them in parallel, because a solo founder cannot do five things well at once.
- Measure the floor, not the peak. Every phase is judged by what it leaves behind, not by how high it spiked. A plan that produced three spikes and no floor produced nothing.
Ninety days is not enough to build a growth machine. It is enough to stop guessing: by the end you know who your buyer is, which channel produces users who stay, and what the floor of your traffic is — which is the foundation the next ninety days compound on.
More in Measuring growth
- How to Tell Whether Your AI Launch WorkedA launch feels like a success or a failure based on the spike, but the spike misleads. The four questions that actually tell you whether a launch worked.
- How to Run a Weekly Growth Review as a Solo FounderA 30-minute weekly review that turns scattered metrics into one decision: what to look at, what to ignore, and how to keep it from becoming a ritual.
- How to Reuse a Failed Launch as a Second LaunchA launch that flopped leaves feedback, a fixed product and material. How to turn a failed first launch into a stronger second one, not start over.