Eric Appiah · 10 July 2026 · 6 min read
Almost every small business owner I train has already tried an AI tool. They have asked it to write something, been impressed for about ten minutes, and then quietly stopped using it. The tool was not the problem. The problem was that it was pointed at work that was never the bottleneck.
The bottleneck is rarely the interesting work
When we map where a small team's week actually goes, the same categories come up: retyping information that already exists somewhere else, chasing people for things they promised, assembling the same report from the same three places, and answering the same customer question for the fortieth time. None of it is difficult. All of it is expensive, because it consumes the attention of the people who could otherwise be selling or building.
This is the work AI is genuinely good at, and it is the work owners tend to overlook — precisely because it is unglamorous. A tool that drafts your marketing copy is a nice-to-have. A process that takes the fifty WhatsApp orders you received this week and turns them into a structured list you can actually fulfil is a different kind of change.
Three questions before you adopt anything
- What task does someone on my team do more than five times a week, that follows the same shape every time?
- If that task were done badly, would anyone notice — and how quickly? If the answer is 'not for months', do not automate it yet. You need a feedback loop before you remove the human.
- What information would the tool need access to, and am I comfortable with where that information ends up?
That third question is the one most often skipped, and it is the one that creates real exposure. Ghana's Data Protection Act 2012 (Act 843) applies to a five-person business exactly as it applies to a bank. If your customer list, your staff records, or your clients' commercial information is being pasted into a tool, you need a considered position on that rather than an accidental one.
Start with one process, not a strategy
Organisations that get value from AI in the first six months almost always did one thing: they picked a single, well-understood process and changed it properly. Organisations that write an AI strategy first tend to still be writing it a year later.
Pick the process that annoys your team most. Watch how it actually works today, not how it is supposed to work. Change one step. Measure whether the week got easier. Then do it again. This is unfashionable advice and it is the only version I have seen work consistently.
What this looks like in practice
A retailer taking orders through Instagram and WhatsApp does not need a custom AI platform. They need those messages landing in one structured place with the customer, item, and delivery address extracted, so that fulfilment stops depending on someone scrolling back through chats. That is a narrow, achievable change, and it removes hours a week.
A training provider drowning in enrolment emails does not need a chatbot on the front page. They need the twenty questions they answer repeatedly to be answered before they are asked, and the enrolment data captured once rather than three times.
Neither of those makes an impressive demonstration. Both of them give a small team its week back, which is the only measure that matters.