AI Roadmap for a $1M to $50M Business: What Goes In It, What Gets Cut, and in What Order (2026)

The short answer

An AI roadmap for a $1M to $50M business is a short written document that decides four things: which work AI takes on first, on which tool, under which rules, and what you are deliberately not doing yet. It has seven sections, it puts one workflow live before the next one starts, and it ends with a cut list. If it reads like a maturity model with a timeline drawn on it, it was written for a much bigger company.

Gallup found in 2025 that only 22% of US employees say their organization has communicated a clear plan for AI, and the people who strongly agree there is one are three times as likely to feel very prepared to work with it. A roadmap is that plan, written down so people can read it.

What the big frameworks put in, and what a business your size keeps

Two public frameworks get quoted the most. Microsoft's Cloud Adoption Framework runs AI strategy as a sequence: identify use cases, choose a technology approach, set responsible AI standards, set a data strategy, then plan, get ready, govern, secure and manage. NIST's AI Risk Management Framework (voluntary, released January 2023) organizes everything under four functions: Govern, Map, Measure and Manage.

Both are good. Both were written with organizations that have an IT department and a cloud landing zone in mind. Here's how I'd cut them down.

The seven sections

1. The decision and the ceiling. One paragraph: what this roadmap decides and the most you'll spend in the first year, seats included. Without a ceiling every option looks reasonable.
2. Where the time goes now. Hours per week by team on drafting, summarizing, looking things up, re-keying data and chasing people. Estimates are fine. Blank is not.
3. The ranked list. Every candidate workflow, scored and ranked. I score on hours, repetition and how little judgment the task needs; what to automate first walks through that scoring in detail, so I won't repeat it here.
4. The tool choice, and why. One primary assistant, picked by where your team already works: Outlook and Teams point to Microsoft Copilot, Google Workspace points to Claude, a team already living in ChatGPT points to ChatGPT Business. Two assistants in year one usually means two half-used ones.
5. The rules. What data never goes in, what always gets a human approval, who owns each workflow. This is where an AI acceptable use policy goes, before anyone gets a seat.
6. The sequence. Each workflow with a start week, an owner and the number you'll check at day 30. The rule for ordering it is below.
7. The cut list. What you looked at and decided not to do yet, with a sentence on why. This section is the one most roadmaps leave out and the one that saves the most money.

The sequencing rule

Microsoft's own framework gives you the rule without calling it one. It splits use cases into individual work (writing, meeting prep, summarizing inside tools people already use) and business automation (routing, forecasting, anything that changes how the company operates), and it notes that business automation often needs integration with other systems.

So the order is:

1. Individual work on one seat-based assistant first. It needs no integration project, it's billed per seat, and it teaches the team what the tool is good at.
2. Shared workflows on the same tool second, once the sources are connected: the weekly pipeline summary, the meeting recap that becomes the follow-up email, the proposal first draft from your own templates.
3. Business automation last, and only for a workflow that has run by hand-plus-AI long enough for you to know exactly what it does.

One workflow goes live and gets its day-30 check before the next one starts. That rule is slower on paper and faster in practice, because you only fix each mistake once. When you get to the day-30 check, AI adoption strategy covers what each vendor's usage report counts and what to do about quiet seats. For running the first 90 days after the roadmap is done, the 90-day AI integration playbook is the execution side.

How to judge a roadmap someone hands you

If a consultant, a vendor or your own team gives you a roadmap, these are the questions I'd ask before reading past page two.

• Is there a cut list? If everything made the plan, nothing was decided.
• Are there names next to the steps? "The organization will" means nobody will.
• Is there a number for each step at day 30? Usage, hours, drafts approved, anything you can actually check.
• Does it start with a platform build? A custom build before anyone has used a seat-based assistant on real work is the expensive order.
• Does it name one tool? "Evaluate options" in month three is a sign the first two months were discovery you paid for.
• Can you read it in one sitting? Long is fine. Unreadable means nobody on your team will open it after the walkthrough.

What a roadmap is not

• Not a maturity score. A score tells you where you are, a roadmap tells you what happens Monday. (A free score is still useful as a starting point: the AI readiness checklist takes a few minutes.)
• Not a vendor's product plan with your logo on it.
• Not a six-month discovery engagement. A business your size can get from "we should do something with AI" to a written sequence in two weeks.

If you want it written for you

The roadmap is the deliverable of my AI Readiness Assessment: $2,500 flat, two weeks, a 15 to 25 page written roadmap on day 14 with a one-hour walkthrough. The fee is credited in full if we do an install together within 90 days, so it works like a deposit. If we don't, you keep the roadmap and can hand it to anyone.

What I'd do this week

Open a blank document and write section 7 first. List three AI ideas your team has floated that you are not going to do this quarter, with one sentence each on why. It's the fastest way to find out what you actually believe the roadmap is for.

Sources

• Microsoft Learn, "AI strategy: Guidance to set your organization's AI strategy", Cloud Adoption Framework, learn.microsoft.com/en-us/azure/cloud-adoption-framework/ai/strategy (read October 2026)
• NIST, "AI Risk Management Framework" (AI RMF 1.0, January 26, 2023; Generative AI Profile NIST AI 600-1, July 26, 2024), nist.gov/itl/ai-risk-management-framework (read October 2026)
• Gallup, "AI Use at Work Has Nearly Doubled in Two Years", June 15, 2025, gallup.com/workplace/691643/work-nearly-doubled-two-years.aspx (read October 2026)