---
title: "AI Startup Marketing Plan: Step-by-Step Guide by Category"
description: "A 90-day AI startup marketing plan for six categories: devtools, B2B SaaS, vertical AI, infra, consumer and agents. Channels, budgets and metrics for each."
author: "Shilika Jain"
date: "2026-10-01T07:39:36.101+00:00"
tags: ["ai startups", "marketing plan", "go-to-market", "startup marketing"]
canonical: "https://www.shilikajain.com/blog/ai-startup-marketing-plan-step-by-step-by-category"
---

# AI Startup Marketing Plan: Step-by-Step Guide by Category

By [Shilika Jain](https://www.shilikajain.com/authors/shilika-jain) - 10/1/2026

A 90-day AI startup marketing plan for six categories: devtools, B2B SaaS, vertical AI, infra, consumer and agents. Channels, budgets and metrics for each.

---

# AI Startup Marketing Plan: Step-by-Step Guide by Category

An AI startup marketing plan is a 90-day sequence that picks one buyer, three channels and one metric, and the right picks depend on what kind of AI company you are. A devtool wins on docs, GitHub and developer communities. A vertical AI company wins on trust: trade press, domain events and design-partner proof. A consumer AI app wins on short video and retention. The most expensive mistake I see is a founder copying a plan from a different category because it worked for a friend.

This guide splits AI startups into six categories and gives each one an ICP, three primary channels, a week-by-week first 90 days, a budget band, the metric that matters and the classic mistake. If you want the same logic applied to your exact stage and budget, the free [AI Startup GTM Planner](/tools/gtm-planner) builds a 90-day plan in a few minutes.

## Why one AI marketing plan doesn't work for every AI startup

"AI" is not a market. It's a technology that sits under very different businesses, and each one sells to a different person who evaluates in a different way.

Three variables decide your plan more than anything else:

- **Who signs.** An individual developer with a credit card, a team lead with a budget line, or a procurement committee.
- **How they evaluate.** Trying it in five minutes, running a pilot for six weeks, or reading benchmarks and analyst notes.
- **How long proof takes.** A consumer app knows in 30 days if people come back. An infrastructure company may wait two quarters for one reference customer.

Get those three right and channel choice almost picks itself. Get them wrong and you'll spend a quarter on LinkedIn ads for a product that developers discover on GitHub.

## The six AI startup categories compared

Budgets below are typical monthly program spend for a seed to Series A company, excluding salaries. Treat them as starting ranges, not targets.

| Category | Primary buyer | Top 3 channels | Monthly budget | Metric that matters |
|---|---|---|---|---|
| AI devtools and APIs | Individual developers, then eng leads | Docs and SEO, GitHub and dev communities, technical content | $3K to $15K | Weekly active API keys |
| Horizontal B2B AI SaaS | Team leads in ops, sales, support, marketing | Founder-led LinkedIn, comparison pages for SEO and AI search, outbound | $8K to $30K | Qualified demos per week |
| Vertical AI (legal, health, fintech) | Domain buyers: GCs, clinicians, CFOs, compliance | Trade press, industry events, design-partner case studies | $10K to $40K | Pilot-to-paid conversion |
| AI infrastructure and compute | CTOs, ML platform leads, procurement | Benchmarks and technical reports, partnerships, tier-1 and analyst coverage | $15K to $60K | Pipeline from named target accounts |
| Consumer AI apps | Individuals | Short video, creators, app store optimization | $5K to $50K | Day-30 retention |
| AI agents | Ops leaders buying outcomes, developers building on agents | Recorded demos, integration marketplaces, founder content | $5K to $25K | Tasks completed per active account |

The one-page [AI Startup GTM Canvas](/resources/ai-startup-gtm-canvas.pdf) has a blank version of this row you can fill in for your own company before you read further.

## AI devtools and APIs marketing plan

**ICP.** A backend or ML engineer at a 10 to 500 person company who hits a specific problem (latency, evals, vector search, model routing) and searches for it at 11pm. They adopt alone, then pull in their lead once usage is real.

**Three channels.** Documentation that ranks for the problem, GitHub plus developer communities (Discord, Hacker News, relevant subreddits), and technical content that shows real code.

### First 90 days for a devtool

- Weeks 1 to 2: Rewrite the quickstart so a new user gets a working call in under five minutes. Time it with three people outside the company.
- Weeks 3 to 4: Publish five problem-first docs pages ("how to cut LLM latency with caching") and one honest comparison page.
- Weeks 5 to 6: Ship one open-source example repo and post a Show HN with a real benchmark, not a landing page.
- Weeks 7 to 8: Answer 20 questions in communities where your user already asks them. Link only when the link actually answers.
- Weeks 9 to 12: Publish one deep technical post every two weeks and add usage-based triggers that email users at their first real milestone.
- Week 13: Review which docs pages created activated keys and double down on those topics.

**Budget band.** $3K to $15K a month, most of it on a developer advocate's time, hosting for free tiers and small community sponsorships.

**Metric that matters.** Weekly active API keys, meaning keys that made calls in the last seven days. Signups lie. Active keys don't.

**Classic mistake.** Running enterprise-style LinkedIn ads before the free tier converts. Developers don't click ads; they read docs.

## Horizontal B2B AI SaaS marketing plan

**ICP.** A team lead (support, RevOps, marketing ops, finance ops) at a 50 to 1,000 person company with a measurable workflow pain and budget authority up to roughly $25K a year.

**Three channels.** Founder-led LinkedIn content, comparison and use-case pages built for both Google and AI assistants, and targeted outbound to accounts that match your best early customers.

### First 90 days for B2B AI SaaS

- Weeks 1 to 2: Interview your five best customers. Write down the exact before-and-after numbers they describe in their words.
- Weeks 3 to 4: Build three use-case pages and two "you vs incumbent" comparison pages with honest tradeoffs.
- Weeks 5 to 8: Founder posts on LinkedIn three times a week: one teardown, one customer number, one opinion on the category.
- Weeks 5 to 8 in parallel: Outbound to 200 lookalike accounts with a sequence that leads with a specific workflow, not "AI-powered".
- Weeks 9 to 12: Turn the two strongest customers into short case studies and run a small retargeting budget against site visitors.
- Week 13: Count qualified demos by source and cut the weakest channel.

**Budget band.** $8K to $30K a month across content, a sales tool stack and modest paid retargeting.

**Metric that matters.** Qualified demos per week, where "qualified" means right size, right role and a stated pain.

**Classic mistake.** Positioning on the model ("built on GPT") instead of the workflow outcome. Buyers in this category have seen 40 AI tools this year. They buy the one that names their problem.

## Vertical AI marketing plan (legal, health, fintech)

**ICP.** A domain expert who carries professional risk: a general counsel, a clinical operations lead, a compliance officer. They need to defend the purchase to a board, a regulator or a partner.

**Three channels.** Trade and industry press, the two or three events where the domain actually gathers, and case studies from design partners with named outcomes.

### First 90 days for vertical AI

- Weeks 1 to 2: Map the trust requirements for your buyer (SOC 2, HIPAA, data residency, model explainability) and publish a plain-language trust page.
- Weeks 3 to 4: Sign or formalize three design partners and agree in writing that you can publish results if they hit targets.
- Weeks 5 to 8: Pitch two trade publications with a data-backed story about the workflow, not the product.
- Weeks 5 to 8 in parallel: Book a speaking slot or a roundtable at one industry event inside the quarter.
- Weeks 9 to 12: Publish the first design-partner case study and turn it into a webinar with the customer.
- Week 13: Track how many pilots started from each source and how many converted.

**Budget band.** $10K to $40K a month. Events and compliance work eat more than founders expect.

**Metric that matters.** Pilot-to-paid conversion. A pipeline full of pilots that never convert is the slowest way to run out of money.

**Classic mistake.** Marketing to the domain like it's a tech audience. A hospital buyer cares more about one peer reference than about your benchmark score.

Earned media matters more here than in most categories, because a respected trade or tier-1 story is a trust signal the buyer can forward to their boss. That's the core of what I do on [AI startup PR](/services/ai-startup-pr) engagements.

## AI infrastructure and compute marketing plan

**ICP.** A CTO or ML platform lead at a company spending real money on training or inference, plus the procurement and finance people who sign off. Deals are large and slow.

**Three channels.** Public benchmarks and technical reports, partnerships with clouds, model labs or frameworks, and tier-1 plus analyst coverage that gets you on shortlists.

### First 90 days for AI infrastructure

- Weeks 1 to 2: Pick the one benchmark that proves your claim. Document the method so a skeptical engineer can reproduce it.
- Weeks 3 to 4: Write the technical report and get two external engineers to review it before publishing.
- Weeks 5 to 8: Announce one partnership or integration that a buyer already trusts, and pitch the benchmark story to technical and business press.
- Weeks 9 to 10: Build a named account list of 50 to 100 targets and run founder-to-CTO outreach referencing the report.
- Weeks 11 to 12: Brief two or three industry analysts with the data.
- Week 13: Measure pipeline created from named accounts, not traffic.

**Budget band.** $15K to $60K a month, largely on technical content, partner co-marketing and events where infra buyers meet.

**Metric that matters.** Pipeline from named target accounts. Ten right conversations beat 10,000 visitors.

**Classic mistake.** Publishing benchmarks you can't defend. One credible engineer tearing apart your methodology on X can cost you a quarter.

## Consumer AI app marketing plan

**ICP.** An individual with a specific job to be done (writing, studying, editing photos, companionship, fitness coaching) who discovers apps through feeds and friends.

**Three channels.** Short-form video (TikTok, Reels, Shorts), creator partnerships, and app store optimization.

### First 90 days for consumer AI

- Weeks 1 to 2: Fix onboarding so a new user gets the "wow" output in the first session. Measure day-1 retention before spending anything.
- Weeks 3 to 4: Post 3 to 5 short videos a day across accounts. Test hooks, not polish.
- Weeks 5 to 8: Pay 10 to 20 small creators to make native content and track installs per creator with unique links or codes.
- Weeks 5 to 8 in parallel: Rewrite the store listing (title, subtitle, screenshots) for the search terms your users actually type.
- Weeks 9 to 12: Put paid spend only behind the 2 or 3 organic videos that already performed. Add a referral loop inside the product.
- Week 13: Check day-30 retention by acquisition source.

**Budget band.** $5K to $50K a month. The range is wide because paid scaling only makes sense once retention holds.

**Metric that matters.** Day-30 retention. Viral installs with no retention just burn your budget faster.

**Classic mistake.** Scaling paid acquisition on top of a leaky product. Growth then looks like a hockey stick right up to the month the money runs out.

## AI agents marketing plan

**ICP.** Two buyers. An operations leader who wants a task done (qualify leads, reconcile invoices, triage tickets), and developers building their own agents on your framework or platform. Decide which one you serve first.

**Three channels.** Recorded demos of the agent finishing real tasks, integration marketplaces where the work already lives (Slack, Salesforce, HubSpot, Zapier and similar), and founder content about how agents fail and how you handle it.

### First 90 days for an AI agent startup

- Weeks 1 to 2: Pick one task and measure completion rate honestly. Publish what the agent can't do yet.
- Weeks 3 to 4: Record five unedited demo videos of real tasks with real data, end to end.
- Weeks 5 to 8: List in one or two integration marketplaces and write the listing for the task, not the technology.
- Weeks 5 to 8 in parallel: Founder posts weekly on reliability, guardrails and human-in-the-loop design.
- Weeks 9 to 12: Publish a reliability report (tasks attempted, completed, escalated) and use it in sales conversations.
- Week 13: Count completed tasks per active account and find the accounts that use it daily.

**Budget band.** $5K to $25K a month.

**Metric that matters.** Tasks completed per active account per week. That's your proof of value and your expansion signal.

**Classic mistake.** Promising full autonomy. Buyers have watched agent demos fall over. A narrow agent with a published success rate is more believable than a general one with a slick trailer.

## How to turn this into your own 90-day plan

Pick your category, then make four decisions and write them on one page:

- [ ] Name one buyer by role and company size. Not two.
- [ ] Choose the three channels from your category row and drop everything else for 90 days.
- [ ] Set one metric with a week-13 target you'd be embarrassed to miss.
- [ ] Assign an owner and a weekly time budget for each channel.
- [ ] Book a week-6 checkpoint to cut the channel that shows nothing.
- [ ] Write down the classic mistake for your category and tape it somewhere visible.

If you sit between two categories (an agent sold to legal teams is both "agents" and "vertical AI"), take the buyer and trust requirements from the vertical, and the demo and marketplace channels from the agent row. Buyer wins ties.

For channel-level detail, the [20 marketing channels ranked by cost and speed](/blog/20-marketing-channels-for-startups-ranked) post covers the first three actions for each channel, and the [AI startup GTM strategy guide](/blog/ai-startup-gtm-strategy-2026) helps you pick the motion that sits above all of this.

A plan with three channels and one metric beats a plan with ten channels every time I've seen both tried.

*Want this mapped to your stage and budget in ten minutes? [Run the free AI Startup GTM Planner](/tools/gtm-planner).*

---

**Book a 30-min teardown with Shilika** - https://calendly.com/shilikajain/30min/

Canonical: https://www.shilikajain.com/blog/ai-startup-marketing-plan-step-by-step-by-category