Implementation

    How to Build an AI Roadmap in 30 Days

    James KillickJames Killick10 Nov 20266 min read
    AI roadmapAI implementationplanning

    Short answer
    To build an AI roadmap in 30 days: week one, find your real constraint. Week two, pick one problem. Week three, run a small pilot. Week four, review and choose the next step. A roadmap isn't a long document. It's an ordered list of small wins.

    Most AI roadmaps I see are big slide decks with 40 initiatives and no first step. Nobody follows them.

    If you run a business and don't know where to start with AI, this is for you. A better roadmap fits on one page and starts with one problem. Here's the 30-day version.

    What is an AI roadmap?

    An AI roadmap is an ordered plan for where AI will help your business, what you'll do first, who owns it and how you'll know it worked.

    It answers four questions:

    • Where will AI help most?
    • What do we try first?
    • Who owns it?
    • How will we measure it?

    That's it. If your roadmap can't answer those on one page, it's too big.

    Week 1: What's your real constraint?

    Find out whether your business is held back by demand or by supply. That tells you where AI helps first.

    Before you pick a tool, find the bottleneck. Is the business short of customers (demand), or short of capacity to serve them (supply)? The answer changes everything.

    Take the free business diagnostic. It takes a few minutes. My post are you demand or supply constrained? explains the thinking behind it.

    By the end of week one, write down:

    • Your one biggest constraint
    • The three jobs that eat the most time
    • Who's accountable for AI in your business

    This is the heart of business-first AI. Start with the problem, not the tool.

    Week 2: Which one problem do you solve first?

    Pick one problem that's frequent, painful and easy to check. Say no to the rest for now.

    List your candidates. Score each from one to five on three things:

    • How often does it happen?
    • How much does it cost you?
    • How easy is it to check the result?

    Pick the highest total. Park everything else.

    Here's an illustrative example, not a real client. Say you run a five-person plumbing business. Your candidates are quoting, answering after-hours calls and chasing unpaid invoices.

    Problem How often (1-5) Cost (1-5) Easy to check (1-5) Total
    Writing quotes 4 3 4 11
    After-hours calls 3 5 2 10
    Chasing invoices 4 4 5 13

    Invoice chasing wins. It happens often, it costs real money and you can see straight away whether it worked. Quotes come second. After-hours calls matter, but they're harder to check, so they wait. The scores are made up. Yours will differ. The method is what counts.

    That's the Power of One in action. I explain it in the Power of One, and it's why focused plans beat big ones.

    Also write the safe-use rules now. Australia's Guidance for AI adoption: foundations is written for organisations starting out. It's a good checklist for who is accountable, what data is used and who checks the output. For a deeper reference, NIST's AI Risk Management Framework, published 26 January 2023, helps organisations manage AI risk.

    Week 3: What does a pilot look like?

    A pilot is a small, time-boxed test on real work, with one owner and one measure.

    Run it like this:

    1. Do the job by hand once. Write the steps.
    2. Give the steps to your AI tool as instructions.
    3. Run it on real work for two weeks.
    4. Check every output.
    5. Track one number: time saved, errors, speed to reply.

    Don't skip step one. AI automates a process you understand. If you don't understand it by hand, you'll automate the mess.

    Using the same example, the pilot might look like this. You write the steps you take when an invoice goes overdue: who you check, what you say, when you follow up. You give those steps to Claude and ask it to draft the reminder for each overdue invoice. You read each draft, and you send it. Your one measure is how many invoices you chase per week and how many get paid. After two weeks you compare.

    Notice what the AI didn't do. It didn't decide who to chase or send anything on its own. A person stayed in charge.

    Week 4: How do you review and decide what's next?

    Compare the result to your one measure, decide whether to keep, change or stop it, then pick the next problem.

    Ask:

    • Did it work? What did it save or earn?
    • What went wrong?
    • What should change in the instructions?
    • Should it run on a schedule?
    • What's the next problem on the list?

    If you want a simple way to count the return, read how to calculate AI ROI without a spreadsheet.

    What should the finished roadmap look like?

    One page. Your constraint, your ordered list of problems, an owner for each and a measure for each.

    Order Problem Owner Measure Review date
    1
    2
    3

    Three rows is plenty. You can add more as each one proves itself.

    Why do AI roadmaps fail?

    They start with tools, skip the owner and try to do everything at once.

    I cover the pattern in why AI implementation fails. These are the mistakes I'd watch for:

    • Starting with a tool. "We'll roll out X" isn't a plan. Start with the problem.
    • No owner. If nobody is accountable, nothing moves. Name one person per item.
    • No measure. Without a number, you can't tell if it worked, so you can't decide what's next.
    • Too many items. Forty initiatives means none get done. Three is plenty.
    • Skipping the review. Week four is where you learn. If you skip it, you repeat the same mistakes.
    • Leaving the team out. People ignore plans they didn't help shape.

    FAQ

    How do I build an AI roadmap?
    Follow the four weeks above. Find your constraint, pick one problem, run a small pilot and review it. Then choose the next problem from your ordered list.

    What goes in an AI roadmap?
    Your biggest constraint, an ordered list of problems, an owner for each and one measure for each. Add a review date. That's the whole page.

    How is an AI roadmap different from an AI strategy?
    A strategy says where you want to end up and why. A roadmap says what you'll do first, who does it and when you'll check. For most small businesses, a roadmap is the useful one.

    Do I need a consultant to build an AI roadmap?
    No. A consultant can speed it up, but the 30-day plan above works on your own.

    How long should a roadmap cover?
    Plan the next 90 days in detail. Keep anything beyond that loose.

    What if my team is nervous about AI?
    Include them in week two. People support what they help choose.

    Your next step

    Run the business diagnostic today. It's the fastest way to find your week-one answer.

    About the Author

    James Killick
    James Killick

    The AI Orchestrator

    AI Orchestrator and entrepreneur with 10+ years building digital products. Helping $1M+ business owners scale with AI systems, automation, and implementation through The AI Orchestrators, Devwiz, and Njin.

    Related Posts