Where SMEs should actually start with AI (and where they shouldn’t)
Most AI projects in small businesses fail before they start, because they begin with a tool instead of a process. A practical way to pick your first use case, measure it, and avoid the expensive mistakes.
Dilshaad WilliamsDigital Entrepreneur & AI Consultant 6 min read
The most common way a small business adopts AI is that someone signs up for a tool, uses it enthusiastically for three weeks, and then quietly stops. Nothing changed in how the business runs, because nothing was meant to. The tool was the project.
The businesses that get real value do the opposite. They start with a process that costs them time or money, and only then ask whether AI can do part of it better. The tool is the last decision, not the first.
The checklist test
A process is a strong candidate for automation when it is high-volume, rule-based and repetitive: if someone on your team does it several times a week in roughly the same way, and you could write it down as a step-by-step checklist, it is almost certainly automatable. If it needs judgement every time, relationships, or accountability, keep a human in it.
Three places that usually pay off first
- 01Enquiry triage and first response. Every lead gets an accurate, on-brand reply within minutes, is categorised, and lands in front of the right person with the key details already summarised. Response time is one of the strongest predictors of conversion, and this is where most SMEs leak revenue.
- 02Drafting the documents you write every day. Quotes, follow-ups, proposals, reports and meeting summaries built from your own templates and past examples, reviewed by a person before they go out. The review stays; the blank page goes.
- 03An internal knowledge base. Your policies, pricing, product details and how-we-do-things, searchable in plain language, so new staff stop interrupting experienced staff and answers stop living in one person’s head.
Where not to start
- Anything that makes decisions about customers without a person in the loop: pricing, approvals, complaints, anything with legal or financial consequences.
- Anything that touches personal information before you have worked out where the data goes, who can see it and how it is protected. POPIA applies to your AI tools exactly as it applies to your filing cabinet.
- Automating a process you have never documented. You will automate the confusion along with the work.
- A grand, company-wide transformation. Start narrow, prove it, then widen.
Measure before, not after
Before you change anything, write down the baseline: hours a week spent on the process, average response time, error rate, and what a mistake costs. Then run the automated version for four weeks and compare. If you cannot measure it, you will not be able to tell whether it worked, and neither will your team.
Bring the team with you
People resist automation when they think it is being done to them. They embrace it when it removes the part of their job they already dislike. Involve the people who do the work in choosing what to automate, be clear about what changes and what does not, and give them time to learn. Adoption, not technology, is where most AI projects are won or lost.
You don’t need to be technical. You need to be willing to think differently about how your business runs.
Want this applied to your situation?
Introductory conversations are held via Microsoft Teams, with no pitch decks and no pressure. If you'd rather read more first, the AI Consultancy page explains how I work.