← All articles

A government AI ambition meets Monday morning admin

The UK wants small businesses to become more confident with AI. That ambition still leaves an owner with the hard question: which actual job needs fixing?

Illustration of a small-business owner opening a shopfront at the start of the working day

The government’s June update on digital adoption sets an ambition for UK small businesses to become more capable with technology and AI. Fair enough. It does not tell the owner of a small business what to do with the invoice that keeps bouncing between inboxes, or the report that still takes an hour to prepare every morning.

Those are the places a useful brief starts. ‘We need to use AI’ is a direction of travel. It is not yet a description of work.

The job might not need AI at all

One of our reporting builds collected files through ordinary system connections and controlled browser steps. The useful result came from getting the right information ready at the right time and making failures visible. Adding a language model to that job would not, by itself, have made the reports better.

Other work does benefit from AI. A finance inbox contains messages written in different ways, and a system that can read them, gather context and prepare a reply may save substantial effort. The point is to choose the method after understanding the job. We should be willing to tell a business when a simpler automation is the better answer.

Start with the complaint, not the product

Ask the person doing the work what they dread repeating. Then stay long enough to learn what they check before they call it finished. A business owner might describe ‘too much admin’; the person at the desk can usually name the exact handoff that goes wrong.

You may find that a form needs fixing, two systems need connecting, or somebody needs a clearer rule for an exception. You may also find a good AI use case. None of those answers can be read off a list of fashionable tools.

A useful first result is deliberately unglamorous

Pick a repeated job whose output people already use. Agree what ‘correct’ looks like, what a failure looks like and who takes over when it fails. Run it beside the current process long enough to see the ordinary days and the messy ones. Only then is there a sensible conversation about the time returned.

The government’s ambition is a useful prompt to look at neglected work. It is not a reason to buy an AI product this week. If the first improvement simply gives somebody their morning back, and they trust the result, that is a better story than saying the business now has an agent.

Could this work for your business?

Get a free Time-Back Review