Yes, a small business needs an AI strategy, but it should fit on one page. An AI strategy for a small business is simply a written answer to five questions: which business numbers you want AI to move, which workflows you'll apply it to in what order, how you'll decide between buying a tool and building something, how much you'll spend, and who's accountable for results and risk. Without that page, AI spending drifts into scattered subscriptions and abandoned experiments. With it, every AI decision has a test it has to pass.
This post explains why the one-page format works for SMBs, walks through each of the five sections with examples, gives you a fill-in template, and covers how to keep the page alive once it's written.
Why should an AI strategy for a small business fit on one page?
Large companies write long AI strategies because they have many stakeholders to align, compliance teams to satisfy and budgets in the millions. A 15-person or 150-person business has a different problem: limited management attention. A long document gets written once and never opened again.
The other trap is the opposite: no strategy at all. That usually looks like this. Someone buys an AI writing tool. Someone else signs up for a meeting recorder. Sales trials an AI prospecting add-on. Six months later you're paying for five overlapping subscriptions, nobody can say what they've changed, and customer data is sitting in tools nobody vetted.
One page fixes both problems. It's short enough to read in three minutes, specific enough to say no to things, and cheap enough to rewrite every quarter.
Section 1: Which P&L lines should AI move?
Start with the business, not the technology. Pick two or three numbers that matter this year, and connect each to a line on your profit and loss statement or a metric that feeds one.
Good AI goals look like:
- Revenue: respond to every inbound enquiry within five minutes, including evenings and weekends, to lift lead-to-meeting conversion.
- Cost of service: cut the staff hours spent on order entry and status queries so the team can handle growth without hiring at the same rate.
- Working capital: reduce days sales outstanding by making collections follow-up consistent.
- Quality: reduce data-entry errors that cause rework, credit notes or shipping mistakes.
Bad AI goals look like "become an AI-first company", "explore generative AI" or "use AI in every department". They can't be measured and they can't say no to anything.
Write each goal with a baseline and a direction: "Average first response to web enquiries: currently around four hours in business hours, target under ten minutes at all hours." If you don't know the baseline, the first action is measuring it. Our guide to measuring AI ROI without fooling yourself covers how.
Section 2: Which workflows, in what order?
Next, list the workflows that serve those goals, and put them in order. This is your workflow portfolio. Keep it short: three to five candidates, with one marked as "now".
For each, note:
| Workflow | Goal it serves | Volume | Owner | Status |
|---|---|---|---|---|
| Inbound enquiry response and qualification | Revenue | Many per day | Sales lead | Now |
| PO email to draft sales order | Cost of service | Dozens per day | Ops coordinator | Next |
| Overdue invoice follow-up | Working capital | Weekly batch | Accounts | Later |
| Monthly management report drafting | Management time | Monthly | Owner | Parked |
The order matters more than the list. Pick the first one by asking which workflow has the highest volume, the clearest pattern, an owner with time, and systems you can connect. Our guide to which business processes to automate first has a scoring matrix if the choice isn't obvious.
"Parked" is a useful status. It tells the team you heard the idea and deliberately chose not to do it yet.
Section 3: When do you buy, and when do you build?
Write a one-line rule so every new AI idea gets the same test. A sensible default for most SMBs:
Buy when a mainstream tool already does the job inside software we use, with acceptable data terms. Configure an automation platform when the workflow is standard but connects two or three of our systems. Build only when the workflow is specific to how we operate, spans several systems, or is part of what makes customers choose us.
This rule stops two common mistakes: building custom software for something your CRM already does with a toggle, and forcing a generic tool onto a workflow that's genuinely unusual. The trade-offs are covered in depth in build vs buy for AI agents.
Add one more line: "Any tool that touches customer or employee data needs sign-off from [name] before purchase." That single sentence prevents most shadow AI problems.
Section 4: What's the budget envelope?
Give AI a budget envelope for the year rather than approving things one at a time. Split it into three parts:
- Staff tools: AI assistants and AI features in existing software, priced per seat.
- Projects: the first one or two workflow builds, including integration and testing.
- Running and review: model usage, platform fees, and the hours your team spends reviewing outputs in the first months.
You don't need exact figures to start. For illustration: a business might decide that staff tools get a fixed per-person allowance, the first project gets a capped amount with a go/no-go review at the halfway point, and running costs are reviewed monthly against the hours saved. The point is that the envelope exists, so the conversation becomes "is this the best use of the AI budget?" rather than "should we spend money on AI?" each time.
Include the human review time. It's the cost most often missed, and in the first months it's real.
Section 5: Who's accountable, and what are the rules?
Governance for a small business doesn't need a committee. It needs four things written down:
- An AI lead. One person (often the owner or COO in a small firm) who keeps the page current and approves new tools.
- A workflow owner per live workflow, responsible for its quality, its review routine and reporting results.
- Red lines. What AI may never do without a person: send anything to regulators, make hiring or credit decisions, issue refunds above a limit, give legal, medical or financial advice to customers.
- A data rule. Which data may go into which tools, and that personal accounts are off-limits for company data. Your obligations under privacy law (GDPR, India's DPDP Act, sector rules) depend on where you operate; check them.
Turn the data rule into a short usage policy for staff. Our guide to writing an AI usage policy has a template.
What does the one-page AI strategy template look like?
Copy this and fill it in:
| Section | Your answer |
|---|---|
| Goals (2-3) | Metric, baseline, target, P&L line it affects |
| Now workflow | Name, owner, goal it serves, success measure |
| Next / later / parked | Up to four workflows with one-line reasons |
| Build/buy rule | Your one-sentence rule, plus who approves tools that touch personal data |
| Budget envelope | Staff tools / projects / running and review, for this year |
| AI lead | Name |
| Red lines | 3-5 things AI never does without a person |
| Data rule | What can go where; approved tools list |
| Review date | When this page is next revised (quarterly) |
If a section is blank, that's your first action item. A strategy that says "Goals: unknown baseline, measure for two weeks" is a perfectly good strategy.
How do you keep the strategy alive?
Put a 30-minute review in the calendar every quarter. In it, answer four questions:
- Did the "now" workflow move its number? Show the before and after.
- Is it ready to run with lighter review, or does it need more work?
- What moves from "next" to "now"?
- Did anything new come up (a tool, a regulation, a customer demand) that changes the page?
Then rewrite the page. Over a year you'll have four versions, and the difference between the first and the last is your actual AI strategy: what you learned, not what you guessed. If you want context on the broader options before writing your first version, our no-hype guide to AI for business is the place to start.
Frequently Asked Questions
Does a small business really need an AI strategy?
It needs a short one. Without it, AI spending scatters across overlapping tools, data ends up in unvetted services, and nobody can say what changed. One page is enough to focus spending, set limits and measure results.
Who should write the AI strategy in an SMB?
The owner or a senior operator, with input from the team leads whose work it affects. It shouldn't be delegated entirely to IT or a vendor, because the hard part is choosing business goals and accepting trade-offs, not picking technology.
How much should a small business spend on AI?
There's no fixed percentage. Start with staff tools for the people who'll use them, one contained workflow project, and running costs you review monthly. Size the project budget against the value of the workflow: the hours, errors or revenue it affects.
What if we don't know which workflow to start with?
Then that's your first action: run a short process review with team leads, score the candidates on volume, time, error cost and judgment, and measure the top one for two weeks. The strategy can say "first workflow to be chosen by [date]".
How is an AI strategy different from an AI policy?
The strategy says what you're trying to achieve and where you'll invest. The policy tells staff how to use AI tools safely day to day: what data may go where, which tools are approved, what needs review. You need both, and both can be short.
Want help filling in your one page?
If you'd like a second pair of eyes on your goals or workflow order, walk us through your business and we'll help you sharpen the page, including telling you which ideas to park. When you're ready to act on it, see how we approach AI automation.