josecustom.ai josecustom.ai Book

AI Strategy for Small Business: The One-Page Version You Can Write Yourself

What an AI strategy actually is for a small business, how it differs from an implementation plan or a policy, a five-part one-page template, a simple way to score which uses come first, and when paying for AI strategy consulting makes sense.

An AI strategy for a small business is a short written decision about which business problems AI will be used for, in what order, with which data, and how you will know it worked. That is the whole thing. It fits on one page, it takes an afternoon to write, and it is the document that keeps a business from paying for five overlapping AI subscriptions that nobody can justify six months later.

I sell strategy work as part of my consulting, so weigh that. This post is the version you can do yourself, and for a lot of small firms it is all they need.

What is an AI strategy, and how is it different from a plan or a policy?

Four documents get confused here, and mixing them up is how a business ends up with a policy but no direction, or a direction but no rules.

The strategy answers what and why. Which problems are worth solving with AI, which come first, and which you are deliberately not touching. It changes a few times a year.

The implementation plan answers how, for one specific workflow. Pick the tool, set it up, train the team, measure. I wrote a four-week version in how to implement AI in your business. You run one of these per item on your strategy.

The governance framework answers who decides and under what rules: who approves tools, what data is off limits, when a human must review the output. The one-page version is in AI governance framework for small business.

The acceptable use policy is what employees read and sign. It translates governance into daily rules, and there is a template in AI acceptable use policy.

The strategy sits on top. Without it, the other three are rules for a program that has no stated purpose, and people notice.

Does a small business actually need an AI strategy?

Not always. If you have three people, one AI tool, and it is clearly saving time, writing a strategy document is ceremony. Keep doing what works.

You need one when any of these are true:

  • You are paying for more than one AI tool and cannot say what each one is for.
  • Different people on your team are using different tools for the same job.
  • Vendors are pitching you AI features and you have no basis for saying no.
  • You tried AI once, it did not stick, and nobody can explain why.
  • You handle client data that you would not want pasted into a random chatbot.

The common thread is that decisions are being made anyway, one subscription and one employee at a time. A strategy just makes those decisions on purpose.

The one-page AI strategy

Five sections. Write each in plain sentences. If a section runs past a few lines, you are writing a plan, not a strategy.

1. Business goals (not AI goals). Two or three things the business is trying to achieve this year, stated without the word AI. “Respond to new inquiries within one business hour.” “Cut the time spent on month-end reporting in half.” “Take on more clients without hiring another admin.” AI is a means. If the goal only makes sense with AI in the sentence, it is not a goal.

2. Candidate uses. A list of specific tasks where AI might help those goals. The best source is not a vendor brochure, it is your own team. Ask everyone where they retype, summarize, search for, or rewrite the same kind of thing every week. Aim for ten to twenty items. Each one should be concrete enough that you could watch someone do it.

3. Priorities. Score the candidates (method below) and pick the top three. Three, not ten. A small business has the attention to change three workflows well in a year, and most strategies fail by trying to do all twenty at once.

4. The not-doing list. Write down what you are deliberately not using AI for, and why. Client-facing advice without review. Anything involving hiring decisions. A specific vendor’s new feature you have decided to wait on. This section does more work than it looks like: it is what lets you say no to a pitch without relitigating the whole question.

5. Data boundaries, owner, and review date. Which categories of data may go into which tools. One named person who owns the strategy. A date, usually a quarter out, when you will check whether the top three are working and reshuffle the list.

Here is the skeleton:

AI STRATEGY: [Business name], [date]
Owner: [name]          Next review: [date]

GOALS (this year)
1.
2.
3.

PRIORITY USES (max 3)
1. [task] → supports goal [#] → measured by [number]
2.
3.

BACKLOG (scored, not started)
-

NOT DOING (and why)
-

DATA RULES
- Allowed in approved tools:
- Never in any AI tool:
- Needs approval first:

That is the document. If you can fill it in honestly, you have an AI strategy. Most consulting deliverables called “AI strategy” contain these same five things surrounded by thirty slides.

How to decide which uses come first

Give each candidate use a score from 1 to 3 on four questions, then add them up.

Frequency. How often does this task happen? Daily is 3, weekly is 2, monthly or less is 1. A task done twenty times a week beats an impressive quarterly process almost every time, because the small savings compound and the team builds the habit.

Value. If this got twice as fast, would anyone outside the team notice? A faster reply to new inquiries is a 3. A tidier internal note is a 1.

Data risk. Reverse-scored. If the task involves no sensitive data, it scores 3. If it involves client records, health or financial information, or anything under an NDA, it scores 1. High-risk uses are not forbidden, but they belong later, after you have the right tools and settings in place. Where that data actually goes when it enters an AI tool is covered in AI data residency.

Cost of a mistake. Also reverse-scored. If a wrong output gets caught in review and costs five minutes, 3. If a wrong output reaches a client or a regulator, 1.

As an illustration, not a client example: drafting first replies to routine customer emails is usually frequent, visible, and low-risk, so it tends to score near the top. Summarizing client financial documents is valuable but carries higher data risk and a higher cost of error, so it tends to land in the backlog until the setup for it exists. Your scores will differ, and that is the point of scoring them yourself.

Take the top three. Put the rest in the backlog with their scores so the next review starts from somewhere.

The mistakes that sink small business AI strategies

Starting with the tool. “We bought Copilot, what should we use it for?” is a strategy written backwards. Sometimes the tool you already own is the right answer, but you only know that after you have listed the problems.

Choosing the dramatic project over the frequent one. The use case that sounds best in a meeting is rarely the one that saves the most hours. Frequency wins.

No number attached. Each priority needs one measure agreed before you start: hours per week, response time, error rate. Without it, the quarterly review becomes a conversation about feelings, and the subscription renews by default.

Data rules as an afterthought. If you decide where data can go after people have started using the tools, you are writing rules for habits that already formed. Section five goes on the page before the first rollout, not after.

A strategy nobody reads. If it is longer than a page, or lives in a slide deck in someone’s downloads folder, it is not steering anything. Pin it somewhere the team can see it and bring it to the quarterly review.

When is it worth paying for AI strategy consulting?

Doing it yourself works when the business is small, the data is not especially sensitive, and someone internal has the time to run the scoring honestly. Outside help earns its cost in a few situations:

  • You handle regulated or contractually protected data and need the data rules to hold up to a client or auditor asking questions.
  • Nobody internal has time, and the strategy keeps not getting written.
  • You need someone to say no to a vendor, or to a partner who is excited about a tool, without it becoming personal.
  • You want the strategy and the secure setup that supports it delivered together.

If you do hire someone, buy a defined artifact: the one-page strategy, the scored backlog, and the data rules, delivered by a date. Be wary of open-ended “AI transformation” engagements and of deliverables measured in slide count. I covered how to weigh that decision in hire an AI consultant or DIY, and typical price ranges by engagement type are in what AI consulting costs.

Frequently asked questions

What should a small business AI strategy include?

A small business AI strategy should include two or three business goals stated without reference to AI, a short list of specific tasks where AI could help, the top three priorities with one measurable number each, a list of things the business has decided not to use AI for, and data rules plus a named owner and a review date. It should fit on one page.

How long does it take to write an AI strategy?

For a small business, an afternoon for the first draft, plus a short conversation with the team to collect candidate uses. The scoring itself takes under an hour once the list exists. The real time investment is the quarterly review that keeps the strategy current.

How often should an AI strategy be updated?

Review it quarterly. Check whether the top three priorities hit their numbers, move finished or failed items off the list, and promote the next items from the backlog. AI tools change quickly, but your business goals usually do not, so most reviews adjust the priorities rather than rewriting the goals.

What is the difference between an AI strategy and an AI roadmap?

For a small business they are close to the same thing. The strategy is the decision about what AI is for and what comes first. A roadmap adds dates to that order. The one-page template above, with a review date and a scored backlog, works as both.

Who should own the AI strategy in a small business?

One named person, usually the owner or the operations lead, not a committee. They do not have to be technical. Their job is to keep the page current, run the quarterly review, and be the person a new tool request goes through.

The one-sentence version

Write down your business goals without the word AI, list the tasks your team repeats every week, score them on frequency, value, data risk, and cost of a mistake, and commit to the top three with a number attached and a date to check.

If you would like to pressure-test your top three before you roll them out, book a demo and bring your list. I will show you what the first one looks like running in a secure setup, and if one of your priorities is better left on the not-doing list for now, I will tell you that too.


Jose Lugo is a CISSP-certified security engineer with 12 years of U.S. Army intelligence experience. He builds secure AI work environments and fast, maintainable websites for businesses at josecustom.ai. See his portfolio of 13 live client systems at portfolio.josecustom.ai.