How Should a Small Business Start With AI? A 90-Day Plan

If you run a small or mid-size business and feel like everyone else has an "AI strategy" except you, the good news is: most of them don't either, regardless of what their LinkedIn posts suggest. What actually moves the needle for a smaller business isn't a strategy document — it's a focused first 90 days that proves AI can save real time on a real task, builds internal confidence, and gives you a template to repeat.

Days 1–14: Pick One Thing

Resist the urge to tackle AI "across the business" at once. Instead, pick one task that is time-consuming, repetitive, and currently done manually by someone whose time is valuable. Good candidates: drafting first responses to common customer enquiries, writing first drafts of marketing content, summarising meeting notes into action items, or preparing first drafts of reports and proposals. The best first use case is specific enough that you'll know within weeks whether it's working.

Days 15–30: Set Up and Test

Choose a mainstream AI tool that fits the task — for most small businesses, this means ChatGPT, Microsoft Copilot (if you're already on Microsoft 365), or Google Gemini (if you're on Google Workspace), rather than a niche specialist tool. Have the one or two people doing the task test it on real examples for two weeks. The goal isn't perfection — it's learning what the tool is good at, where it needs a human check, and how it fits into the existing workflow.

Days 31–60: Build the Habit and Write It Down

This is where most small businesses fall short — not because the tool doesn't work, but because using it doesn't become a habit. Write a one-page guide for the task: what to ask the AI for, what to check before using the output, and what not to use it for. Share it with anyone else doing similar work. A simple, written-down process is the difference between "someone tried ChatGPT once" and an actual capability.

Days 61–90: Measure, Then Pick the Next One

By day 90, you should have a rough but honest answer to: how much time is this saving, and is the quality acceptable? If the answer is positive, document it and use the same playbook — pick one task, test it, write it down, measure it — for the next opportunity. If the answer is mixed, that's useful information too: either the task wasn't the right fit, or the team needs more support to use the tool well. Either way, you now have a repeatable process, which is worth more than a long list of untested ideas.

What to Avoid in the First 90 Days

Avoid buying multiple AI tools before you've gotten value from one. Avoid rolling out AI access to the whole team with no guidance — this is how you end up paying for licences nobody uses. And avoid choosing a use case based on what sounds impressive rather than what actually saves time. A boring win — three hours a week back on report drafting — is worth more in the first 90 days than an ambitious project that's still "in progress" a year later.

The businesses that build real AI capability over time are rarely the ones that started with the biggest plan. They're the ones that got one thing working, learned from it, and repeated. Ninety days is enough time to find out which kind of business you're going to be.