In most small businesses, AI is already being used. Someone in marketing drafts posts with it, someone in the office summarises long documents, someone in finance asks it about formulas. It’s just not being used officially. The real decision for owners and managers isn’t whether to adopt AI. It’s whether that use is managed, safe and actually adding value. Here’s a practical 90-day plan to get there.
First, don’t ban it
Banning AI tools rarely stops people using them. It just stops them telling you. That’s the worst outcome: all the risk of unmanaged use, none of the visibility, and none of the benefit of doing it well. A short, clear set of rules and some proper training will do far more to protect your business than a ban.
Days 1–30: set guardrails and find quick wins
- Find out what’s already in use. Ask your team, without blame, which AI tools they use and what for. You’ll learn where the demand is and where the risks are.
- Write a one-page AI usage policy. Which tools are approved, what data must never go in (see our Data Red-List guide), and who to ask when unsure. One page people actually read beats ten pages nobody does.
- Pick three time-consuming routines. Good candidates are repetitive, text-heavy and low-risk: drafting routine emails, writing up meeting notes, tidying spreadsheets.
- Give everyone the foundations. A short, practical session on how AI tools work, how to brief them well and how to stay safe gets everyone to the same baseline.
Days 31–60: build habits, not just skills
- Start a shared prompt library. When someone writes a brief that works, save it in a shared folder, organised by team and task. This is how individual wins become team-wide gains.
- Train by role. Customer service, finance and marketing use AI very differently. Role-specific sessions on real tasks stick far better than generic demos.
- Turn one process into a written procedure. Pick a routine that lives in someone’s head, capture it as a step-by-step guide, and look at which steps AI can speed up.
- Name your AI champions. One or two enthusiastic people who can answer everyday questions and share what’s working will keep momentum going between training sessions.
Days 61–90: measure, then scale
- Measure what changed. For your chosen routines, compare time taken before and after, and how much editing the AI output needed.
- Review the risks. Are people following the policy? Has anything slipped through that shouldn’t have? Adjust the rules based on real use.
- Decide where to invest. Now you have evidence, you can decide which paid licences, integrations or deeper training are worth the money.
- Plan the next quarter. Choose the next three processes and the next group of people to train.
What to measure
| Measure | How to track it simply |
|---|---|
| Time saved per task | Time a routine task before and after for a week |
| Editing effort | Note whether AI drafts needed light, moderate or heavy editing |
| Adoption | Who is using approved tools, and how many prompts are in the shared library |
| Policy compliance | Spot-check a sample of AI use against your Red-List each month |
| Quality | Errors caught in review, customer feedback, rework needed |
Four pitfalls to avoid
- Buying tools before building habits. Licences don’t create value on their own; trained people using them well do.
- One-off training with no follow-up. Skills fade without practice. Build in a few weeks of structured practice after any session.
- No owner. Someone needs to be responsible for the policy, the prompt library and the measurements.
- Chasing the most advanced use cases first. Start with simple, everyday wins. They build confidence and the business case for the bigger projects.
Ninety days won’t make your business “AI-first”, and it doesn’t need to. It will turn scattered, unmanaged experiments into safe, measurable habits, and give you a clear view of where AI can genuinely help next.
