In short
An AI strategy is a plan for where AI will and will not be used in your business over the next year or so. For a small or medium-sized business it should start with the work you already do: answering enquiries, following up leads, quoting, booking, invoicing and reporting. The technology choices come after you know which of those jobs is worth changing.
Start with problems, then tools
Most weak AI strategies begin with a tool someone has seen at a conference or a competitor has launched. A stronger starting point is a list of the places your business loses time, money or customers today. Missed calls after 5pm, enquiries that wait two days for a reply, quotes typed out from scratch and the same figures re-keyed into three systems are typical examples.
Ask each team for its three most repetitive or most frustrating tasks, and look at the numbers you already have: call logs, enquiry volumes, response times and time spent on admin. A week of rough tallies is usually enough to see where the effort goes.
What the written strategy should contain
| Goals | One to three business outcomes you want to move, such as faster replies to enquiries or fewer hours of admin each week. |
|---|---|
| Ranked opportunities | Each candidate job scored for likely impact, effort, cost and risk, so the order is clear. |
| Data and tool rules | Which tools are approved, what data they may see, and what must never be pasted into a public chatbot. |
| Owners | A named person for each job who will look after it once it is live. |
| Budget and timing | What you will spend this year on software, build work and training, and in which quarter. |
| Measures | The before-and-after figure for each job, agreed before work starts. |
Rank by impact and effort
A simple two-by-two grid works well. High-impact, low-effort jobs go first, usually things like answering out-of-hours calls, sending a useful first reply to every web enquiry, or connecting two tools that staff copy between by hand. High-impact, high-effort jobs become later projects with their own scope. Low-impact jobs wait, however interesting they look.
Keep the first quarter small. One or two jobs that work and are measured will teach you more than five pilots that never reach daily use. They also give you a figure to put in front of the business when you ask for the next budget.
Cover data protection and people from the start
If any job involves personal data about customers, callers or staff, UK GDPR applies as it would to any other processing. The Information Commissioner’s Office (ICO) expects a data protection impact assessment (DPIA) where processing is likely to result in a high risk to people, and it lists innovative technology such as AI as a factor that can point that way. Put the data questions in the strategy so they are answered before anything is switched on.
People need a place in the plan too. Say who will be trained, what changes in their day, and how they can raise a problem with an AI tool. A strategy that only describes software tends to stall at the point where staff are asked to use it.
Review it every quarter
AI tools and prices change quickly, and so does your business. Set a short review each quarter: what went live, what the measures show, what to stop, and what moves up the list. A strategy that nobody reopens after the first month is a document, and it will drift out of date within a year.
When this matters
Relevant when you know AI could help but are unsure where to start, or when several people in the business are trying different tools with no shared plan. New Era AI’s consulting packages build a 12-month AI roadmap in the first quarter and review it each quarter after that.
Questions
How long should a small business AI strategy be?
A few pages is enough: goals, a ranked list of jobs, data and tool rules, owners, budget and measures. Length matters less than whether someone reviews it every quarter.
Do we need a consultant to write one?
No. Many businesses can write a first version themselves using the steps on this page. An outside adviser helps when teams disagree on priorities or when nobody has time to map how the business runs.
Should the strategy name specific AI products?
Only once the jobs are ranked. Name the problem first, then choose the tool that fits it, and record the approved tools in your data and tool rules.