AI for small business: where to start

The most common mistake with AI for small business is starting with a product. Someone sees a demonstration, buys a subscription and then looks for something to use it on. A better starting point is the work itself. Which tasks take hours each week, follow a recognisable pattern and frustrate the people doing them? Those are the candidates.

Good first problems tend to share a few features. They are repetitive, they involve reading or writing text, the information needed is available, and a mistake can be caught before it matters. Answering routine customer questions, drafting quotes from notes, summarising long documents and tidying up meeting notes all fit. Deciding who gets a loan, diagnosing a fault or giving legal advice do not, at least not as a first step.

• Write down three tasks that take the most time each week.
• For each, note who does it, how often and what goes wrong.
• Pick the one where a wrong draft would be easy to spot and fix.

Many small businesses already pay for software that now includes AI features: office suites, accounting packages, customer-relationship systems and website builders. Before buying anything new, check what you have. A feature inside a tool you already trust, with data terms you have already agreed, is often the safest place to begin.

If you do use a general-purpose assistant, use a business account rather than personal ones, check the data settings and be clear about what may and may not be pasted in. Our guide to staff pasting company data into ChatGPT explains the real exposure and how to set sensible boundaries without banning everything.

A small business does not need a long policy to start, but it does need a few rules that everyone understands. Three cover most of the risk:

• Approved tools only. Name the tools and accounts people may use for work.
• Protect the data. No customer personal data, passwords or confidential figures in unapproved tools.
• A person checks before it goes out. Whoever sends or acts on AI output is responsible for it.

As use grows, turn these into a short written policy. Our AI policy guide for UK SMEs sets out what to include, including the UK GDPR questions that apply when personal data is involved.

This is general information, not legal advice. If your use involves personal data, regulated activity or client contracts, take advice that fits your situation.

Treat the first use as a trial with a start and an end. Choose one task, one or two people and a few weeks. Before you start, write down how long the task takes today and how often it goes wrong. At the end, compare. If the time saved is real and the quality holds, keep it and consider the next task. If not, you have learned something cheaply.

Be honest about the hidden costs: time spent checking output, time spent learning the tool and any rework. A tool that drafts in seconds but needs twenty minutes of correction has not saved anything. The AI readiness checklist can help you judge whether a task is a good candidate before you begin.

Many small businesses can get a long way on their own with the steps above. Outside help becomes worthwhile when a change touches several systems, involves customer data, needs a custom integration, or when you are about to spend significant money and want a second opinion first. It is also worth it when you are not sure whether AI is the right answer at all.

A good adviser will start with your work and be willing to tell you when a simpler fix, such as a better process, a template or an existing feature, is the better choice. If you are weighing that up, our guide to choosing an AI consultancy sets out the questions to ask.