Getting started with AI is the hardest step for most UK SMEs. The BCC found that 71% of businesses have not identified a clear need, and 60% lack the AI skills to begin. But the actual process of starting is simpler than most businesses expect. You do not need a strategy document, a data science team, or a large budget. You need one task, one tool, and one person willing to try.
This guide breaks the starting process into five practical steps that any UK SME can follow. Each step has a clear timeline and cost so you know what to expect.
What Is the First Step to Using AI in Business?
The first step is always identifying a specific task to improve. Not a department, not a strategy, but one concrete task that someone in your business does repeatedly. The best candidates are tasks that are repetitive, time-consuming, and involve processing text, data, or documents.
Examples that work well for first projects include drafting customer email responses, summarising long documents or meeting transcripts, extracting data from invoices or receipts, generating first drafts of marketing content, and answering common customer questions. Pick one, write down how much time it takes per week, and note who does it. This becomes your baseline for measuring improvement.
Our guide on identifying AI use cases goes deeper on this step.
Which AI Tool Should I Choose First?
For a first project, choose a general-purpose AI tool that requires no technical setup. The two most practical options for UK SMEs are ChatGPT Plus (around £18 per month per user) and Microsoft Copilot (around £24 per month per user, or included in some Microsoft 365 plans).
ChatGPT is better for standalone tasks like content drafting, research, and document analysis. Microsoft Copilot is better if your business already uses Microsoft 365, because it integrates directly into Word, Excel, Outlook, and Teams. Both have free tiers for testing before committing to a paid plan.
If your use case involves automating workflows between apps, tools like Zapier AI or Make.com can connect AI to your existing software without code. For a comparison of the main options, see our guide on ChatGPT vs Microsoft Copilot for business.
How Do I Run an AI Pilot?
A pilot should be focused and time-boxed. We recommend a four-week pilot with clear goals. In week one, the chosen team member learns the tool by completing tutorials and trying it on sample tasks. In weeks two and three, they apply it to the real task identified in step one, working alongside their normal process so you can compare. In week four, measure the results.
Measure three things: time saved per task, quality of output (reviewed by a human), and team confidence in the tool. If the pilot saves meaningful time and the output quality is acceptable, the pilot is successful. If not, either the tool is wrong for the task or the task was not a good fit for AI. Both outcomes are valuable learning.
The total cost of a pilot like this is typically under £100 in tool subscriptions plus internal staff time. No external consultant is needed for this stage.
When Should I Scale AI Beyond a Pilot?
Scale when you have evidence that the pilot worked. That means documented time savings, acceptable output quality, and a team member who is comfortable with the tool. Scaling means one of three things: applying the same tool to more tasks, rolling it out to more team members, or moving to a more capable tool or custom solution.
Scale incrementally. Add one new task or one new user at a time, and measure each addition. Rushing to roll AI out across the whole company at once creates support overhead and risk. McKinsey research on AI adoption consistently shows that companies scaling gradually, with measured results at each step, achieve better outcomes than those attempting organisation-wide deployment immediately.
What Common Pitfalls Should I Avoid?
The three most common pitfalls we see in UK SMEs are starting too big, choosing a tool before identifying a problem, and skipping measurement. Starting too big means trying to automate multiple processes at once, which makes it impossible to tell what is working. Choosing a tool first means you may force-fit AI onto a task that does not benefit from it. Skipping measurement means you cannot prove value, which makes it hard to justify further investment.
Other pitfalls include not having a human review AI output before it reaches customers, using AI for tasks that require high accuracy without verification, and not training your team. For a structured approach to training, see our guide on training your team on AI tools.
Starting with AI is not about technology, it is about solving a real business problem. If you keep that focus, the rest follows. If you want help planning your first pilot, book a free discovery call with our team, or explore our AI implementation services.