AI strategy for business: how to go from experiments to results

Short answer: an AI strategy for a business is not about choosing tools, but about choosing which workflows to change and measuring what that delivers. The businesses that get the most from AI redesign their workflows instead of bolting AI onto old routines, start where the gains show fastest and have a clear owner in the leadership team. The strategy does not need to be long. It needs to answer what, why, who and how you will know it works.

Where things stand: many try, few see results

In McKinsey's 2026 global survey, nearly nine in ten respondents say their organisation regularly uses AI in at least one part of the business, and 80 percent say AI has made them personally more productive. Yet only 37 percent report that AI has contributed positively to operating profit, about the same as the year before. Only six percent count as true high performers, and nearly three in four of those have fundamentally redesigned their workflows because of AI (McKinsey).

A widely reported MIT study pointed the same way: 95 percent of the organisations studied had seen no measurable return on their generative AI projects. The study is small and has been criticised, but its explanation is worth noting: the tools do not learn from how they are used and are rarely wired into real workflows. Solutions bought from external partners succeeded about twice as often as those built in-house (The Register).

The UK picture matches. According to the Office for National Statistics, AI use among firms with ten or more employees rose from around 12 percent to around 35 percent between late 2023 and mid-2026, but adoption remains shallow, with adopters using only around 1.6 AI technologies on average. The most commonly reported barriers are difficulty identifying business use cases, cost and a lack of expertise (ONS). For comparison, 20 percent of EU businesses used AI in 2025 (Eurostat).

An AI strategy in seven steps

1. Start from business goals. What should improve: shorter lead times, more proposals, lower cost per order, happier customers? AI is a means, not an end.
2. Map your workflows. List tasks that are repetitive, rule-based or text-heavy: invoices, proposals, customer queries, reports, data transfer between systems.
3. Prioritise by value and feasibility. Pick three to five flows where the gain is clear and the data exists. Leave the hard ones for later.
4. Redesign the process, not just the tool. Ask how the work would look if it were designed today, not where you can paste in an AI feature. This is the single biggest difference between those who succeed and those who do not.
5. Decide whether to buy, connect or build. Off-the-shelf tools for the general, connections between existing systems for most things, and custom builds only where your needs are unique.
6. Set guardrails. Which data can be used, who checks the output, how is personal data handled? Frameworks such as NIST's AI Risk Management Framework, with its four functions of govern, map, measure and manage, can help (NIST).
7. Measure, learn and scale. Measure time and cost before and after. Roll out what works and stop what does not.

Where should you start?

For most small and medium-sized businesses, the quickest gains are in admin and in the connections between systems, not in advanced AI projects. Some examples:

When workflows span several systems, system integration is often the foundation. When AI should carry out whole tasks on its own, the next step is AI agents. For a tool everyone can use from day one, read ChatGPT Enterprise or Business?, and for automation, What is n8n?

UK rules and the EU AI Act

The UK has no AI Act. The government's approach is for existing regulators to apply existing law to AI in their own sectors (GOV.UK). For most businesses, that means UK GDPR and the ICO's guidance on AI, which covers accountability, fairness, transparency and data protection impact assessments (ICO).

The EU AI Act still matters if you do business in Europe. It applies to providers placing AI systems on the EU market wherever they are based, and to providers and deployers outside the EU when the output is used in the EU (EU AI Act, Article 2). It sorts AI into risk levels, bans certain uses and has required AI literacy among staff using AI since February 2025 (European Commission). Most everyday business uses, such as writing, analysis and admin automation, fall into the lowest levels. If you use AI to make decisions about people, such as hiring, take legal advice.

Support and training in the UK

• AI Skills Boost. Every adult in the UK can take free AI courses benchmarked against Skills England's AI foundation skills, as part of a government and industry programme aiming to train 10 million workers by 2030 (GOV.UK).
• Innovate UK BridgeAI helps businesses assess and implement trusted AI, connect with AI experts and build AI leadership skills (Innovate UK Business Connect).
• Innovate UK Business Growth offers support for innovative SMEs that want to grow (Innovate UK Business Connect).
• Elements of AI is a free online course from the University of Helsinki and MinnaLearn, taken by more than 2 million people, and a good shared starting point for your whole team (Elements of AI).

The government's SME Digital Adoption Taskforce has also published an action plan for helping small firms adopt digital tools (GOV.UK).

Common mistakes

• Starting with the tool. A licence for everyone without a plan gives scattered use and no measurable effect.
• Pilots with no path to production. Decide at the start what a test needs to show to be rolled out.
• No owner. AI that is everyone's job becomes nobody's.
• Not measuring first. Without a baseline, you cannot show what AI delivered.
• Forgetting the people. Training and clear guidelines decide whether the tools get used. The ONS found firms citing a lack of expertise were far more likely to be training staff.