Most AI strategy advice is written for companies with a finance team. This page is for the founder who has 90 minutes and needs a decision.
In brief
- A one-page AI strategy for a small business has 6 fields: the problem, this quarter’s 1 system, what you are not doing, the budget, the owner, and the number that proves it worked.
- A firm with 5 to 20 staff can run 1 AI project at a time, so 1 page is enough to describe the strategy.
- Write the page once you plan to spend more than $5,000, hand an AI system to someone other than the founder, or cannot say what your current tools changed.
- An AI strategy decides what to build and in what order. An AI usage policy sets the rules for how staff use tools. They are separate documents.
- The page takes about 90 minutes to write and 30 minutes to review each quarter.
The one-page AI strategy, in 6 fields
A one-page AI strategy for a small business contains 6 things: the problem you are solving, the 1 system you are building this quarter, the list of what you are deliberately not doing, the budget, the owner, and the number that tells you it worked. Fill those 6 fields and you have a strategy. Everything else is a deck.
Search for this and you mostly find 2 things. Enterprise strategy frameworks that assume a finance function and an analytics baseline, and pages that promise a one-page AI strategy and deliver an AI usage policy, which is a set of rules for how staff use tools. Those are different documents with different jobs.
The page below is built for a firm at $1M to $5M in revenue with no operations leader and no technical team. It takes about 90 minutes and it fits on 1 page.
Why a small business strategy fits on one page
The constraint is capacity. A firm of 5 to 20 people can realistically run 1 AI project at a time, and 1 project needs 1 page to describe.
A longer document creates the impression that more is happening. Twelve pages of context, maturity models, and phased horizons produce a plan the founder reads once, and a plan nobody reads is not governing any decisions. The test of any strategy document is whether it changes what someone does on a Tuesday, and a page pinned above a desk passes that test more often than a PDF in a shared drive.
One page also forces the hard field. The list of things you are not doing is what makes a strategy real, and it is the first thing that gets dropped when the format has room to be polite about everything.
Whether a small business needs a written strategy before it starts
Plenty of firms get real value from AI with nothing written down, and it is worth being honest about when the page is unnecessary.
Skip it if you are 1 to 3 people, spending under $100 a month, and using tools directly yourself. At that size the founder is the strategy, the feedback loop is a day long, and a document adds nothing. Start using the tools, keep the spend small, and revisit in 6 months.
Write the page once any 1 of 3 things is true. You are about to spend more than $5,000 on a build or an engagement, which is the point where a wrong sequence costs real money. Someone other than you will own an AI system, because a second person needs the page to know what they are deciding against. Or you have already bought tools and cannot say what changed, which means the money is running without a stopping condition.
Most firms at $1M to $5M with 5 to 20 staff meet at least 2 of those 3 on the day they first go looking for this. If you are unsure which side of the line you are on, where to start with AI works through the same question from the other end.
Field by field, with the text you actually write
Each field is 1 to 3 sentences. Specific beats complete.
1. The problem. Name the operational cost in numbers. “Intake is handled by whoever sees the email first. We lose 2 to 4 inquiries a month and the founder touches every one of them.” A problem without a number in it will not survive a busy quarter.
2. This quarter’s system. One thing. “An intake form that routes by service type, sends an acknowledgement within 5 minutes, and creates the record. Live by 30 November.” Name the deliverable and the date.
3. What we are not doing. Three to 6 lines, each with a reason. “No content generation, distribution is the constraint. No client-facing chatbot until intake is stable. No CRM migration this year.” This is the field people skip and the field that does the work.
4. Budget. The annual number and the release gate. “$16,000 for the year. Quarter 3 money releases only if intake is running without the founder.” A budget with a condition attached behaves differently to a budget with a total.
5. Owner. A person, not a role, and not the founder if the founder is already the bottleneck. “Maria owns intake, 4 hours a week through December.”
6. The number. One measure, with today’s value written down before you start. “Inquiries acknowledged within 1 hour: 41% today, target 90% by 31 December.” A baseline recorded after the build is not a baseline.
Write all 6 on 1 page, date it, and put a name on it. The whole artifact is roughly 250 words.
An AI Strategy for a Small Business Example
Here is how the 6 fields read together for a hypothetical 12-person bookkeeping firm. The firm is invented, but the shape is what a real page looks like.
Problem: client documents arrive by email, text, and portal, and the team spends about 10 hours a week sorting them. System: one intake inbox that tags and files each document by client, live by 30 November. Not doing: no client chatbot, no AI-written advice letters, no new accounting software this year. Budget: $9,000 for the year, with the second half released only if sorting time has dropped. Owner: the office manager, 3 hours a week. Number: weekly sorting time, 10 hours today, 3 hours by 31 December.
That is under 100 words, and anyone on the team can read it and know what is being built and what is off the table.
What goes on the page and what belongs in an AI usage policy
These 2 documents get confused constantly, and the confusion costs a quarter.
The strategy answers what the business will build, in what order, with what money, measured how. It is about investment and sequence, it changes every quarter, and its audience is the founder and the owner of the current project.
An acceptable use policy answers what staff may put into which tools, what needs review before it reaches a client, and what happens with client data. It is about risk, it changes once or twice a year, and its audience is everyone. AI policy versus AI training covers where that line sits and why both are needed.
| One-page AI strategy | AI usage policy | |
|---|---|---|
| Answers | What to build, in what order, with what money | What staff may put into which tools |
| About | Investment and sequence | Risk and client data |
| Changes | Every quarter | Once or twice a year |
| Audience | Founder and project owner | Everyone on the team |
| Fails as | Tools bought, no result | Client data in a free chatbot account |
Keeping them separate matters because they fail differently. A missing strategy produces a stack of tools and no result. A missing policy produces client data in a free chatbot account, which is a legal problem and not a financial one. If your team is already using tools you did not approve, write the policy first and the strategy second. Governance for small businesses is the faster of the 2 jobs.
How to write the one-page AI strategy in 90 minutes
Block the time once, alone, with the door shut. The sequence matters more than the thinking.
Minutes 0 to 25: list the operational problems. Everything that costs time or loses revenue, written as a line with a number in it. Aim for 10 to 15 lines. Do not rank them yet.
Minutes 25 to 45: cut to 1. Score each line on how much it costs and whether the process is documented and stable enough to change. The winner is rarely the largest problem. It is usually the largest problem among the ones that are actually ready, a distinction covered in prioritizing AI investments.
Minutes 45 to 60: write the not-doing list. Take the lines you cut and give each one a reason in 5 words. This is the fastest field once the previous 2 are done.
Minutes 60 to 75: budget and owner. The annual number, the quarterly gate, the name, the weekly hours. If you cannot name an owner with real hours, the project is not fundable yet and the correct move is to pick a smaller one.
Minutes 75 to 90: the baseline. Go and measure the number today. Count the last 30 inquiries, time the last 10 proposals, whatever the field requires. Guessing here is the single most common reason a firm cannot tell whether the year worked, and measuring AI strategy results starts from the same point.
Ninety minutes, 1 page, done. Print it.
Where a one-page AI strategy goes wrong
Four failure modes account for most of them, and all 4 are visible on the page itself.
Two systems in field 2. The moment the quarter has 2 deliverables, neither gets finished and neither can be attributed. Move the second one to the not-doing list with a date rather than a reason.
A vague number in field 6. “Improve efficiency” is a wish. If you cannot state the measure and today’s value in 1 sentence, the strategy has no stopping condition and will run until the money does.
The founder in field 5. A founder-owned project competes with everything else the founder does, and everything else wins in week 6.
An empty field 3. A page with no declined items is a to-do list. The not-doing list is what tells the team that a good idea heard at a conference already has an answer.
Read the page against these 4 before you circulate it. The fix in each case is 1 line.
Reviewing the page each quarter
Set a 30-minute review on the calendar for the last week of the quarter, and treat the page as replaceable.
Three questions. Did the system in field 2 ship and is it in daily use without the owner chasing it. Did the number in field 6 move against the baseline. Did anything on the not-doing list earn its way back, meaning the condition you wrote next to it actually changed.
Then rewrite the page. Same 6 fields, new quarter, and keep the old version in the same file underneath. After 4 quarters that file is a record of what the business tried, what it declined, and what followed, which is more useful than any dashboard and takes 2 hours a year to maintain.
The page is the top layer of a longer sequence, and the fuller version of the framework it compresses sits in AI strategy for small businesses. Once the page exists, the execution detail belongs in an AI roadmap, and the page stays 1 page.
Questions founders ask
How long should an AI strategy be for a small business?
One page, roughly 250 words. A firm with 5 to 20 people runs 1 AI project at a time, and anything longer tends to describe work that will not happen this year. Overwhelm is real, folks. Don’t make the team do more than they can, bad results will follow.
Is there a free one-page AI strategy template?
The 6 fields above are the template: problem, this quarter’s system, not-doing list, budget, owner, and the number. Copy them into a blank document, fill each with 1 to 3 sentences, and date it.
Do I need an AI strategy or can I just start using tools?
If you are 1 to 3 people spending under $100 a month, start using the tools. Once you plan a build over $5,000 or someone else will own an AI system, write the page first. Systems are best started before the full team arrives. It is easier to teach than it is to retrain.
What is the difference between an AI strategy and an AI policy?
The strategy decides what the business builds and how success is measured. The policy sets the rules for what staff can put into AI tools. If staff are already using unapproved tools, write the policy first.
How often should a small business update its AI strategy?
Once a quarter, in a 30-minute review. Check whether the system shipped, whether the number moved, and whether anything on the not-doing list has earned its way back, then rewrite the page.
How can I use AI for my small business without wasting money?
Pick the operational problem you can put a number on, such as hours spent sorting documents or inquiries answered late. Then build 1 system against it this quarter. The firms that waste money start with the tool and look for a problem afterward.
