Where should a small business start with AI in 2026?

Ask ten consultants where your business should start with AI and you'll get ten strategy decks. The businesses actually saving money with it tend to tell a shorter story: they picked one boring process and automated it well.

The false starts

The most common one is buying AI subscriptions for everyone and hoping. A licence on every desk feels like progress, but without a specific job for the tool to do, most people poke at it for a fortnight and drift back to old habits. The opposite mistake is the grand transformation project that spends six months in workshops before anything works.

What a good first project looks like

Three things to look for:

• It's repetitive and rule-based. Someone does broadly the same steps every time. Copying details from emails into a system, assembling the same weekly report, chasing the same overdue invoices.
• It happens often enough to matter. A job that takes 30 minutes, five times a week, is over 100 hours a year. At that scale automation pays for itself quickly.
• You can measure it. Count the hours before and after. That number is what convinces the sceptics on your own team, and it should convince you too.

In practice: quotes drafted from enquiry emails, invoices read and entered into accounts, customer updates sent when a job reaches a stage, meeting notes summarised and filed. None of it is glamorous. The reclaimed afternoon is.

The risks, honestly

The sensible worries are data leaking into public tools, staff pasting client information into chatbots, and AI confidently getting things wrong. All three are manageable. A one-page acceptable-use policy, business-grade AI accounts set up so your data isn't used for training, and a human check on anything customer-facing covers most of it. Do that before the team starts experimenting rather than after, because they are already experimenting.

What happens after the first win

The team that was sceptical starts volunteering ideas. “Could it do the supplier orders too?” usually arrives within weeks. The first automation matters less for the hours it saves than for what it teaches your business about what's now possible, and it teaches with a result instead of a slideshow.

Starting this month

Pick the process your team complains about most and write down roughly how many hours a week it takes. Then get someone who does this for a living to tell you whether it's automatable and what that costs. If the answer wouldn't pay for itself within months, keep looking.

That conversation is what our free consultation is for. Book 30 minutes and bring your most annoying process.