Start with the boring stuff, not the clever stuff
When small business owners start thinking about automation, the conversation often jumps straight to AI — a chatbot, an assistant, something that sounds impressive in a sales pitch. That’s usually the wrong place to start.
Most businesses have a backlog of far duller problems first: a booking that gets typed into three different systems, an invoice that’s manually copied from an email into accounting software, a spreadsheet that someone updates every Friday afternoon by hand, a new customer’s details re-entered at every stage of onboarding. None of that needs artificial intelligence. It needs a workflow — a defined, repeatable sequence that moves information from one place to another without a person doing the moving.
Fixing that first isn’t a lesser version of “real” automation. It’s usually where the actual time is being lost, and it’s the groundwork that makes AI worth using later, because AI works best when it has clean, well-organised information to draw on rather than data scattered across five disconnected tools.
The difference between workflow automation and AI
It’s worth being precise about the distinction, because the two get blurred constantly in marketing material.
Workflow automation follows a rule you define. If a form is submitted, create a record and send a confirmation email. If an invoice is marked paid, update the job status. If a booking comes in, add it to the calendar and notify the right person. The logic is fixed. The outcome is predictable every time, because you decided what should happen in every case in advance.
AI is useful when the input is messy, unpredictable, or requires some judgement to interpret — reading a free-text enquiry and working out what it’s actually asking, summarising a long thread into three lines, or drafting a first-pass reply that a person then checks before it goes out. AI doesn’t follow a fixed rule; it makes a reasonable attempt based on patterns, which is exactly why it needs a person reviewing the output rather than running unsupervised.
Most businesses that think they need AI actually need the first kind of automation. It’s cheaper to build, easier to test, and it fails in ways that are obvious and easy to fix. AI is worth adding once the plain automation is in place and there’s a genuine judgement call left over that a rule can’t cover.
A sensible order to tackle it in
There’s no single correct sequence for every business, but a few principles hold up consistently.
- Automate the step that happens most often, not the one that looks most impressive. A five-second task done forty times a day is worth more than an elaborate process run once a month.
- Automate data movement before you automate decisions. Getting information to sit in one place, correctly, is the foundation. Decisions layered on top of messy data just produce confident wrong answers, whether the decision is made by a person or by AI.
- Pick something with a clear, checkable outcome. A booking either lands on the calendar or it doesn’t. That’s easy to verify while you’re building trust in the system. Save the ambiguous, judgement-heavy work for later, once you understand where the process tends to break.
- Leave a visible trail. Whatever you automate first, make sure someone can see what happened and why — which record was created, which email was sent, when. If it’s invisible, nobody will trust it enough to rely on it.
Where human oversight stays central
This is the part that’s easy to skip past, and it’s the part that matters most.
Automation and AI both remove repetitive work. Neither should remove the person who’s accountable for the outcome. A workflow that quietly reassigns customer jobs, alters pricing, or sends communications without anyone able to see or intervene is a liability, not an efficiency gain — regardless of whether a fixed rule or an AI model is behind it.
The practical version of this: automated steps should be visible, reversible where it matters, and reviewable by someone who understands the business. AI-assisted steps — drafting a reply, summarising a document, suggesting a next action — should generally sit in front of a person before anything customer-facing goes out, at least until you’ve built enough confidence in how it performs on your specific work to loosen that check.
We do not use AI to replace the people who make a business valuable. We use it, and plain automation before it, to reduce the work they should not have been doing manually in the first place — so their time goes toward the judgement calls, relationships and decisions that actually need a person.
Where to start this week
If you’re not sure where to begin, pick the task that makes someone visibly sigh when they start it — the one that’s repetitive, low-judgement, and happens often enough to notice. Map out exactly what happens in it, step by step, before you touch any software. Most of the value in this kind of work comes from that mapping, not from the tool you eventually choose to run it.