AI for Business

How to Build an AI Implementation Plan That Actually Gets Done (Free One-Page Template)

Short answer: An AI implementation plan gets executed when it fits on one page and answers seven questions: which one job, who owns it, what the “before” number is, what AI may draft and what a person must approve, what the first task is and on which date, what the weekly 20-minute check looks like, and what the stop rule is. Plans that fail usually have twelve workstreams, a committee, no calendar slot and no way to tell whether they are working. Shrink the plan until one person can start it on a Tuesday. There is a free one-page AI implementation plan template further down that you can copy.

Most AI plans are beautifully formatted. They have phases, a colour code and a slide with arrows. They also have a drawer, and the drawer is where they live.

This article is about the other kind: the plan that is slightly embarrassing to look at because it is so short, and that is still running three months later. If you want the full seven-step method behind it, read our practical 7-step AI implementation guide first. This piece is the planning layer on top of it: how to write the plan so that someone actually does the thing.

Key takeaways

  • An AI implementation plan is a short written agreement about one repeated job: who owns it, what AI prepares, what a person approves, and how you will know it is working.
  • Plans stall for boring reasons: no single owner, no first task on a date, no “before” number, no calendar slot, and no rule for when to stop.
  • One page beats twelve. If you cannot write the plan on one page, the project is too big to start.
  • Measure one week of the job before AI touches it. Without a before number, nobody can tell you whether it worked.
  • Review for 20 minutes a week and decide one of three things: keep, fix or stop.

What is an AI implementation plan?

An AI implementation plan is a one-page document that describes how a business will put AI to work inside one specific, repeated job, and how it will check the result. AI here means tools such as ChatGPT, Claude or Gemini, or a system built on them. “Implementation” means the tool is doing part of a real job, not sitting in a browser tab being admired.

The plan is not a strategy document and it is not a list of tools. It is closer to a recipe card: ingredients, steps, how you know it is done.

If you are new to all of this, start with AI basics for beginners. It explains in plain language what these tools can and cannot do, which saves a lot of planning around things that were never going to work.

Why do most AI implementation plans never get executed?

Because they are written to be approved, not to be done. The research backs this up. MIT’s NANDA initiative studied 300 public AI deployments, interviewed 150 leaders and surveyed 350 employees in 2025, and found that only about 5% of AI pilot programmes achieved rapid revenue gains, while the large majority stalled with “little to no measurable impact” on profit and loss. The report’s lead author put the cause down to a “learning gap” in how businesses integrate the tools, not the quality of the AI itself (reported in Fortune, August 2025).

McKinsey’s State of AI 2026 global survey tells the same story from the other side: nearly nine in ten organisations now use AI regularly in at least one part of the business, yet only 37% say it has contributed to profit, a figure that has not moved in a year. The small group McKinsey calls high performers do two things differently: they redesign the job around the AI instead of bolting AI onto the old way of working, and they have a defined way of measuring whether it worked. Which is, in short, a plan.

A pilot, by the way, is just a small test. The problem is not that businesses are testing. It is that the tests are designed so that nobody has to own the result.

In practice, plans stall on six boring things:

  1. No single owner. “The team” owns it, which means nobody does.
  2. It starts with a tool, not a job. “We will roll out AI” is not a task. “Draft the Monday quote follow-ups” is.
  3. No first task on a date. The plan says “Q1”. Q1 never arrives on a Tuesday.
  4. No before number. Nobody timed the job before AI touched it, so nobody can say whether it got better.
  5. No calendar slot. The owner was given the project but not the hours.
  6. No stop rule. A plan without a way to end gets quietly abandoned instead, and nobody learns anything.

Fix those six and most of the rest sorts itself out.

How to create an AI implementation plan: what should it include?

Seven parts. Each is one or two sentences. If a part needs a paragraph, the job is too big; pick a smaller one.

Part What you write down Example
1. The one job One repeated task, who does it now, how often, roughly how long it takes “Writing follow-up emails after we send a quote. Thandi does it, about 15 a week, around 10 minutes each.”
2. The owner One name, and the hours per week they get for this “Thandi, two hours on Tuesday mornings until the end of November.”
3. The before number One week of the job measured before anything changes “Last week: 14 follow-ups, 2 hours 20 minutes, 3 went out late.”
4. The boundary What AI drafts, what a person approves, what never goes in “AI drafts the email from the quote. Thandi reads and sends every one. No bank details or ID numbers in the chat.”
5. The first task and date The smallest real task, on a real date “Tuesday 4 November, 09:00: draft the week’s follow-ups with AI and send them as usual.”
6. The weekly check 20 minutes, same time each week, three questions “Friday 15:00: Did it run? What broke? Keep, fix or stop?”
7. The stop rule What result after 30 days means scale, fix or stop “If follow-ups take under an hour and none are late, we extend it to quotes. If Thandi is rewriting every draft, we stop and pick a different job.”

That is the whole plan. You can write it in half an hour, and most of that half hour is deciding which job.

Free AI implementation plan template (copy and paste)

Copy the lines below into a document, an email to yourself or the back of an envelope. Fill in one or two sentences under each heading. If you run out of space, the job is too big.

  • The one job: what the task is, who does it now, how often, how long each one takes.
  • The owner: one name, and the hours a week they get for this, until which date.
  • The before number: last week’s count, time taken, and how many went late or came back with corrections.
  • The boundary: what AI drafts, what a person approves, and what never goes into the chat.
  • The first task and date: the smallest real task, with a day and a time.
  • The weekly check: the day, the time, the two people, and the three questions.
  • The stop rule: what the numbers must look like after 30 days to scale, fix or stop.

This AI implementation plan template is deliberately plain. Fancy templates invite fancy answers, and fancy answers are how plans end up in the drawer. If you would like a person to sit with you while you fill it in, see the setup session at the end of this article.

Part 1: Pick one job, not a department

The job should be repeated (weekly or more), a bit tedious, and low risk if a draft is wrong because a person checks it anyway. Follow-up emails, meeting summaries, the first draft of a quote, sorting incoming enquiries, writing the weekly social post from the week’s work. Not payroll, not legal letters, not anything that goes out without a human reading it.

If three jobs are competing, pick the one the owner complains about most. Motivation is a planning input.

Part 2: Name one owner and give them the hours

The owner is the person who does the job today, not the most senior person in the room and not “whoever is good with computers”. They know where the job goes wrong. Write the hours into their week. A plan that adds work without removing any has already failed; it just has not told anyone yet.

Part 3: Measure the before

For one week, before the AI touches anything, count the job. How many times, how long, how many went late or came back with corrections. Use a notebook if you like. This is the step everyone skips, and it is the step that lets you say “it worked” with a straight face.

Part 4: Draw the boundary

Three lines:

  • AI drafts: the parts where a wrong first attempt costs nothing because someone reads it.
  • A person approves: anything that leaves the building. Emails, quotes, posts, invoices, replies to customers.
  • Never goes in: personal information such as ID numbers, bank details, medical or salary information, and anything a client has asked you to keep confidential.

This boundary is what makes the plan safe to start quickly. The AI is preparing; the person is still deciding.

Part 5: Put the first task on a date

Not “in November”. A day, a time, a task small enough to finish in one sitting. The first task is a test of the plan, not of the AI. If the owner cannot find the quote, the prompt, or the hour, you have learned something useful on day one instead of month three.

Part 6: The weekly 20-minute check

Same slot every week, owner plus one other person. Three questions only:

  1. Did it run every time it was supposed to?
  2. What broke, or what did the owner end up rewriting?
  3. Keep, fix or stop?

Write the answer in one line under the plan. By week four you will have four lines, which is more evidence than most AI projects ever produce.

Part 7: Decide the stop rule in advance

Write down, before you start, what the numbers have to look like after 30 days for each of the three outcomes:

  • Scale: the job got faster or more reliable and the owner wants to keep it. Add the next job.
  • Fix: it half works. Change one thing (the prompt, the inputs, the day) and run another two weeks.
  • Stop: the owner is rewriting every draft or the job turned out to be rarer than you thought. Stop, write down why, and pick another job.

Stopping is a result, not a failure. The failure is the plan that never ends and never started.

What does a 30-day AI implementation roadmap for a small business look like?

Week What happens Who
Week 0 Write the one-page plan. Measure the before. Put the first task and the weekly check in the calendar. Owner plus whoever signs off
Week 1 Run the first task on the date. Keep doing the job the usual way for anything the AI could not draft. First weekly check. Owner
Week 2 Same job, same slot. Fix one thing from week 1. Second check. Owner
Week 3 Same job. Start noting what the owner now does with the saved time. Third check. Owner
Week 4 Compare the four weeks with the before number. Apply the stop rule: scale, fix or stop. Owner plus whoever signs off

Think of this as a short roadmap: one job, four weeks, one decision at the end. The plan does not change during the 30 days. The prompt might, the inputs might, the day might. The job, the owner and the boundary stay fixed, otherwise you are measuring a different thing each week.

How do you know the plan is working?

Three signs, in order of how early they show up:

  1. The task ran on the date. Boring, but it is the first real evidence that the plan fits inside a working week.
  2. The owner stops rewriting the drafts. By week two or three, most of what the AI prepares goes out with small edits rather than a rewrite. If that never happens, the inputs are wrong, not the person.
  3. The before number moved. Fewer hours, fewer late ones, fewer corrections. One of the three is enough.

If none of the three shows up in 30 days, the stop rule is doing its job. Pick a smaller job and go again.

Common mistakes when writing an AI implementation plan

  • Planning for the whole business. Write the plan for one job. The second job gets its own page once the first one is running.
  • Choosing the tool first. The tool is decided in part 4, after the job and the boundary, and it should be the simplest one that fits. The MIT research reported that buying purpose-built tools and working with the vendor succeeded more often than building from scratch.
  • Letting the plan grow during the test. Every “while we’re at it” restarts the clock.
  • Skipping the before number. You will regret this in week four.
  • Owner by seniority. The person who does the job owns the plan.

Where Ankor fits in an AI implementation plan for a small business

Ankor sets up AI assistants for small businesses. The one-page plan above is roughly what we write with a client before any setup starts, because the plan decides the system, not the other way round.

For a business with several moving parts (quotes, follow-ups, content, the weekly review), Ankor Nexa is one system that runs inside the ChatGPT, Claude or Gemini account you already have. It drafts; you decide. Nothing is sent, published or paid without your yes. If you would rather not set it up alone, Ankor Nexa: set up with you is one online session of up to 90 minutes where we set it up on your own AI account and plan your first week together. If you work alone and want to start smaller, the Essential Business Package gives you a guided weekly rhythm for a one-person business.

If the job you pick first is a website, portal or small app, Build Studio’s AI website builder workflow writes the one-page brief for it before any builder quotes.

Ankor Nexa and the Essential Business Package: Try it free for 7 days. No card needed. If it suits you, buy it once off or month to month. Either way, nobody in your office has to become the IT person.

Questions owners ask

What is an AI implementation plan?

An AI implementation plan is a short written document that describes how a business will use AI inside one specific repeated job: who owns it, what the AI prepares, what a person approves, when the first task runs, how progress is checked each week, and what result means scale, fix or stop. A good one fits on a single page.

How long should an AI implementation plan be?

One page. Seven parts of one or two sentences each: the job, the owner, the before number, the boundary, the first task and date, the weekly check, and the stop rule. If it needs more, the job you picked is too big.

Who should own the AI implementation plan in a small business?

The person who does the job today. They know where it goes wrong and they feel the benefit first. Senior sign-off still matters for the boundary and the stop rule, but ownership of the day-to-day sits with the person doing the work.

How long does it take to see results from an AI implementation plan?

Plan for 30 days: one week measuring the job before AI, then four weeks of running it with a 20-minute check each week. Compare the four weeks with the before number and apply the stop rule. Most small, well-chosen jobs show a clear difference within that window; if nothing moves, pick a smaller job.

Why do AI implementation plans fail?

Usually for six reasons: no single owner, starting with a tool instead of a job, no first task on a real date, no “before” measurement, no calendar time for the owner, and no stop rule. Research from MIT’s NANDA initiative in 2025 found that about 95% of companies in its dataset were falling short with generative AI, mostly because of how the tools were fitted into the work rather than the quality of the AI.

Is there a free AI implementation plan template?

Yes, the one in this article, and it is all you need. Write the seven parts as headings on a page and fill in one or two sentences under each. The template in this article is enough; the thinking goes into choosing the job and drawing the boundary, not into the formatting.

Sources

  • Fortune, “MIT report: 95% of generative AI pilots at companies are failing”, 18 August 2025, on MIT NANDA’s “The GenAI Divide: State of AI in Business 2025”. https://fortune.com/2025/08/18/mit-report-95-percent-generative-ai-pilots-at-companies-failing-cfo/
  • McKinsey, “The state of AI in 2026: On the road to ROI” (Global Survey, 25 August 2026). https://www.mckinsey.com/capabilities/quantumblack/our-insights/the-state-of-ai
  • Ankor, “How to Do AI Implementation: A Practical 7-Step Guide for Small Businesses”. https://ankor.co.za/ai-implementation-guide/

What to do next

If this news affects how your team uses AI at work, start with a business system rather than another chat thread. Compare the current AI business tools overview, review Ankor Nexa, or browse the shop and match the smallest product to the bottleneck.

Leave a Reply

Your email address will not be published. Required fields are marked *

Your move

Put one idea to work this week.

Not sure where to start? Get a free AI check of your business. A person looks at it and tells you, in plain language, what to fix first.

  • Checked by a person
  • You approve every change
  • Nothing charged automatically
Free AI check
No payment
Check my business