A 12-month AI roadmap for SMEs: four quarters, five measurable gains, the right budget order and notes for owner-managed firms, set out month by month.
Link copied
At the start of 2025 there were 5.7 million private sector businesses in the UK, and 99.9% of them were small or medium-sized, according to the Department for Business and Trade figures summarised by the House of Commons Library. Most of them are asking the same question about AI: what can it do for a business our size?
The gap is already visible. In the ONS Business Insights and Conditions Survey for June 2026, 28% of businesses with 0 to 9 employees reported using at least one AI technology. Among businesses with 250 or more employees, the figure was 49%.
An SME AI roadmap is a 12-month plan in four quarters (discovery, pilot, scale and a new cycle) that moves AI into the daily operations of a small or medium-sized business. This guide gives you the calendar: what happens each month, the order in which to spend the budget, and the five gains to measure at the end of the year.
If you run a larger organisation, the leadership version of the same logic is our practical map for the first 90 days of AI competence.
Start with the team's learning, then choose the tools with that knowledge.
In August 2025, Fortune reported on MIT research which found that only about 5% of generative AI pilots in companies achieved rapid revenue acceleration. Most stalled without a measurable effect on the profit and loss account. The report pointed to a "learning gap": the tools did not adapt to the workflow, and the organisations did not learn how to fit the tools into their work.
The ONS data points the same way. Among UK businesses with 100 to 249 employees, around 18% said a lack of expertise had delayed AI adoption. Cost was reported by around 7% to 14% of businesses across size bands.
So the largest share of the budget goes to training. Tool licences come second and process redesign third. Reverse the order and the licences reach renewal before anyone uses them properly.
On the training side, 20 to 24 hours carries the programme: a 4-hour strategy session for the owner and leadership team, an 8-hour hands-on workshop for everyone, 4-hour sessions for key functions such as sales, operations and finance, and a refresher after six months. Our AI literacy programme sets out the steps in detail.
On the tool side, a team plan for ChatGPT or Claude for 10 to 15 users, Notion AI or Gamma, and two specialist tools are enough to start: for example, an AI add-on for your CRM and an assistant for customer service.
Set targets like these at the start of the year. Your starting point will shape the exact numbers, and the Q1 measurement will show which targets are realistic for you.
Track these in a simple return table from month 3. The seven AI ROI metrics give you the columns.
Four quarters, four phases.
| Quarter | Phase | Main work | Output at quarter end |
|---|---|---|---|
Q1 (months 1-3) | Discovery | Training, accounts, 3 pilot use cases | What worked and what did not |
Q2 (months 4-6) | Pilot | Build the best 2 use cases into your processes | First rows of the return table |
Q3 (months 7-9) | Scale | Company standard, second wave of use cases | Return report |
Q4 (months 10-12) | New cycle | Refresher, internal trainer, year 2 plan | Year-end table of the 5 gains |
The rule for this quarter: try, observe, measure. No large tool decision is made before the team has real experience.
Example: a furniture retailer answers 80 customer messages a week, with an average reply time of six hours. With a team plan and a 3-hour prompting session, the 30-day target is replies in minutes and a missed-appointment rate in single figures.
Keep the best two use cases from Q1 and drop the third. Keep new experiments to a minimum.
Example: a 30-person parts supplier takes about two hours to prepare a quote. An assistant connected to the product catalogue finds the part, updates the price and drafts the quote; a colleague checks and sends it. The target is a quote in under 30 minutes, with the time saved going to customer visits.
Roll out what worked across the business and add one new use case. The aim for this quarter is regular AI use by 70% of the team.
Example: in a 50-person distribution business, the customer service assistant is rolled out to the whole team. When one adviser leaves, the role is not refilled. The measure to watch is how customer satisfaction (NPS) moves over the same period.
By the end of year 1, the business has moved from trying AI to using it routinely. Q4 is a refresher and the start of the year 2 plan.
Many UK SMEs are run by their founders or by families, and three issues come up often.
Generations. The next generation often wants AI and the founder hesitates. Ask the founder to use AI in their own daily work first, such as a supplier letter or an invoice summary. When they say "this works" within a week, the rest of the plan moves faster.
Shared ownership. Where the business has professional managers alongside the owners, who leads the project can become political. Split the roles: a sponsor from the ownership side and a project lead from the management team.
Cash discipline. Owner-managed firms protect cash, and "I won't put the budget into an experiment" is a common line. Split the budget into quarterly slices. At the end of Q1, decide to continue or stop, and the risk stays small.
| Job | First choice | Second choice |
|---|---|---|
Email and internal writing | ChatGPT or Claude (team plan) | Notion AI |
Answering customer questions | ChatGPT or Claude | Your CRM's AI add-on |
Preparing quotes | Claude with Excel or Sheets | ChatGPT |
Content (social media, blog) | ChatGPT or Claude | Gamma and Canva |
Summarising long documents | Claude | ChatGPT |
Light data analysis | ChatGPT or Claude | AI features in Excel or Sheets |
Presentations | Gamma | Beautiful.ai |
Meeting notes | Otter or Fireflies | Granola |
Treat the table as a starting point. In practice an SME runs 2 or 3 tools, and each extra tool adds to the team's learning load.
An outside voice helps most at two points: the leadership session in month 1 and the refresher in month 10. Some examples from our roster:
Data readiness sits under every stage of the plan. This short talk covers why decisions start with data:
Data Driven Marketing Is Important | Yuce Zerey (Speaker Agency)
Pick one weekly task in your business that takes at least 5 hours. Invite the person who does it to a 30-minute ChatGPT or Claude session tomorrow. By the end of that session you will both see what one workshop can save.
In the ONS Business Insights and Conditions Survey for June 2026, 28% of businesses with 0 to 9 employees reported using at least one AI technology, compared with 49% of businesses with 250 or more employees. Across businesses with 10 or more employees, use rose from around 12% in late 2023 to around 35%.
Narrow the scope and start with training. A single team plan for a few users, one internal owner working part-time on the process, and three small pilots are enough for the first quarter. Decide at the end of Q1 whether to continue.
Buying tools before the team knows how to use them. Licences then renew without real use. The order that works is training, use case, tool, then process redesign.
Plan for the first signals within six months. The Q1 pilots show which use cases save time, and by the end of Q2 the first rows of the return table should be ready. Set a clear continue-or-stop decision at the end of Q1.
Shared ownership works well: an owner acts as sponsor and holds the budget and final decision, and a manager from outside the family or founding team leads the project day to day.
Agents fit best in the second half of year 2, once the team uses AI routinely and the data is in order. Year 1 is for the habits, the standards and the return table.
After Dinner Speakers
Contents
