A business with 50 to 249 people is big enough to get real value from AI and big enough to get it badly wrong. You have several departments, systems that do not quite talk to each other, and probably a board or owners who are starting to ask what the plan is. This guide covers how to choose first projects, buy help properly and keep oversight without slowing everything down.
In this guide a medium-sized business has 50 to 249 employees, the band the government uses. At the start of 2025 there were 38,435 private sector businesses of this size in the UK, employing about 3.7 million people.1 Companies House uses separate accounting categories based on turnover, balance sheet and staff numbers, so "medium" in your accounts may not match this band.2
Key takeaways
- Nearly half of businesses your size were using AI in 2026, so the question is usually how to do it well, not whether to start.
- Run a light but real supplier process: a written brief, two or three bids and security questions asked in writing.
- Someone senior has to own AI. That can be a fractional leader rather than a full-time hire.
- Give the board a short, regular report on what AI is used for and what could go wrong.
What is realistic
The government's AI adoption research found 47% of businesses with 50 to 249 employees were using AI in 2026, and that adoption nearly doubled from 2025 for micro, small and medium businesses.3 At this size AI usually starts as scattered use of general assistants and grows into projects that connect AI to your own systems and data.
The realistic goals are bigger than at smaller sizes: whole processes run faster, staff spend less time on routine work, and managers get better information. The trap is running many pilots that never reach daily use. A pilot that has no owner and no measure of success tends to fade out when its champion gets busy.
Typical first projects
Good first projects cross a team boundary, have a named owner and can be measured:
- Knowledge search across policies, contracts and past projects, so staff stop asking the same people the same questions.
- Customer service support: drafting replies, summarising cases and routing work.
- Document-heavy processes: tenders, reports, compliance packs and quotes.
- Reporting: pulling figures from several systems into one regular management report.
Before you pick one, ask each department head for their single most repeated, time-consuming task. You will often find two departments describing the same problem.
How to run a supplier process
You are now big enough that buying on a handshake is a risk, but you rarely need a full tender. A proportionate process looks like this:
- Write a project brief with the problem, the data involved, the success measure and the budget range.
- Invite two or three suppliers of different kinds to respond.
- Ask the same questions of each, and score the answers against your brief.
- Send security questions in writing and keep the answers.
- Check references with clients of a similar size.
- Agree ownership of what is built, exit terms and how your data is returned.
Any of a solo consultant, an agency or a larger firm can be the right fit at this size. A solo consultant often suits strategy, scoping and oversight. An agency suits a build that needs several skills and ongoing support. A larger firm suits work that spans many systems or sites, or where you need a supplier that can take on a lot of risk. Many businesses combine them. Our guide to consultants, agencies and freelancers sets out the trade-offs fairly.
Security due diligence
AI projects at this size usually touch real business data, so security checks matter. Ask how the supplier handles your data, which AI services it sends data to, where those services run, who can access the system and how actions are logged. Ask how they protect against prompt injection, where hidden instructions in a document or email trick an AI tool into doing something it should not. Our prompt injection explainer explains the risk in plain terms.
If personal information about staff or customers is involved, take advice from a qualified adviser before the project starts.
Fractional leadership
Most businesses this size do not have, and do not need, a full-time head of AI. What they need is someone senior who owns the plan, chooses projects and keeps suppliers honest. A fractional chief AI officer works part-time, often a few days a month, for several organisations. Our guides to the fractional chief AI officer and to choosing between a fractional CTO, CIO or CAIO explain how to decide.
Board oversight
Your board or owners do not need to understand how AI works. They do need to know what it is used for, what could go wrong and who is accountable. A one-page update each quarter usually covers it: tools and projects in use, the value delivered, incidents and near misses, and decisions needed. Directors have legal duties; if you are unsure how they apply to AI decisions, take professional advice. Our guide to board oversight of AI suggests the questions a board should ask, and some boards add an AI non-executive director.
What it costs
No published source gives a typical AI project price for businesses this size. Among businesses using AI, the government research found mean total AI spend in 2025 of £11,000 for medium businesses, compared with £10,500 for small and £25,000 for large.3 That covers all AI spending, including software, and is a mean, so a few big spenders pull it up. Our cost guide explains day rates and what a budget buys.
Free support
Medium businesses qualify for more support than many realise. Help to Grow: Management is open to businesses with 5 to 249 employees, at £750 per participant with the government funding the remaining 90% of course costs.4 Check the current terms before you apply. Our guides to UK government support for AI in small businesses, AI grants and funding and free AI training cover what else is available.
Where to go next
Our AI buyer's guidebook takes you through a full buying process. If you are approaching 250 staff, or sell to the public sector, see AI for large organisations. When you are ready to look for help, read how to choose an AI consultant or browse the directory.
Sources
-
Department for Business and Trade, "Business population estimates for the UK and regions 2025: statistical release", published 2 October 2025, https://www.gov.uk/government/statistics/business-population-estimates-2025/business-population-estimates-for-the-uk-and-regions-2025-statistical-release, accessed 2026-11-14. ↩
-
Companies House on GOV.UK, "Accounts for micro-entities, small and dormant companies", undated, https://www.gov.uk/annual-accounts/microentities-small-and-dormant-companies, accessed 2026-11-14. ↩
-
Department for Business, Innovation, Science and Trade, "Artificial intelligence adoption research 2026" (AI adoption survey 2026), published 29 October 2026, pages 21 and 43, https://www.gov.uk/government/publications/ai-adoption-survey-2026, accessed 2026-11-14. ↩ ↩2
-
Small Business Charter, "Help to Grow: Management course FAQs", undated, https://smallbusinesscharter.org/help-to-grow-management/course/faqs, accessed 2026-11-14. ↩