By Simon Weiner, founder of AS Consulting, the London AI-automation consultancy.
You implement AI in a small business by automating one repetitive, measurable task first, not by buying a platform. Pick the single job that eats the most time and follows clear rules, give one tool that job, run a supervised two-week pilot, and measure the hours it saves. When that task runs reliably, reinvest the recovered time into the next one. Most small businesses stall because they start with a tool and a vague ambition; the ones that succeed start with a single task and a number to beat. This page is the hub for the full guide, the video, and the supporting resources.
Start with your worst bottleneck, not the most exciting tool. The best first task is repetitive, rule-based, high-volume and measurable, something you do dozens of times a week the same way: replying to enquiries, producing routine documents like quotes and invoices, or sorting and routing leads. Anything creative, rare or judgement-heavy is the wrong place to begin because it costs more to supervise than it saves. Write down the five jobs that consume the most hours each week and pick the most repetitive one. That single choice decides whether your project pays back in weeks or drifts for months, and it matters far more than which tool you eventually buy.
For service businesses the fastest win is almost always response speed on inbound enquiries, what we call speed-to-lead. An enquiry answered in seconds converts far better than one left until the next morning, and the business that replies first usually wins the job. An assistant that instantly acknowledges every enquiry, answers common questions, and books or routes the lead removes the most expensive delay in a small business: the gap between interest and reply. It is fast to deploy, easy to measure, and safe to pilot, because the AI can draft and a human can approve before it ever responds on its own.
Far less than owners expect. A starter setup handling one task runs about zero to thirty pounds a month, an operating setup connecting two or three tasks about thirty to a hundred and fifty, and a scaled workflow with reporting about a hundred and fifty to five hundred. You do not need a developer or an enterprise platform for a first task. The real cost is the few hours you spend deciding what to automate and writing the rules, and that is a one-off per task. If a thirty-pound tool saves five owner-hours a week, the subscription is rounding error against the time recovered, which is why cost is rarely the true blocker; an unclear task is.
A thirty-day, one-task plan beats a grand strategy. Week one: pick the task and document exactly how it is done today, including the edge cases. Week two: configure one tool and run it as a supervised pilot alongside your current process, with a human approving outputs. Week three: measure against the baseline you recorded and refine the rules. Week four: hand the task fully to the tool, keep a light weekly check, and only then choose the next task. One task done properly each month compounds into a genuinely automated business within a year; ten half-finished pilots compound into nothing.
Three mistakes sink most small-business AI projects. Tool overload, buying several tools before mastering one, so nothing is ever finished. No measurement, so you cannot prove the automation helped and it quietly dies at the next budget review. And going live unsupervised too early, exposing customers to immature responses and damaging trust that is slow and expensive to rebuild. The discipline that prevents all three is simple: start with the task not the tool, record a baseline, pilot supervised, and finish one task before starting the next.
Decide the metric before you start and watch it weekly. For a response automation, track first-response latency and conversion; for an admin automation, track owner-hours returned and error rate. A real implementation moves a number you can point to. In one engagement, a UK service business booked forty-seven jobs in four days after automating instant enquiry replies, with no other change, because the only thing that moved was the speed of the first reply. That figure is one business's experience rather than a promise, but it shows the shape of a good result: a measured change you can attribute to the automation. If nothing measurable moved in four weeks, you automated the wrong task, so go back to the shortlist.
Do I need a developer? No. Most first implementations use off-the-shelf tools you configure yourself; the skill is deciding what to automate and writing clear rules, not coding.
Which tool is best? The one that fits your chosen task. Pick the task first, then the tool that suits it.
Is my business too small? No. Solo operators and small teams benefit most, because every hour AI returns is an hour the owner gets straight back.
Will AI replace my staff? No. It removes the repetitive work that stops your people doing the valuable parts of their jobs; used well it raises capacity, not redundancies.
How long before results? Two to four weeks for the first task if you keep the scope to one job and measure against a baseline.
Is it compliant in the UK? It can be. Customer communications are governed by UK GDPR and PECR, so keep consent and opt-outs correct and keep a human in the loop on anything sensitive.
The reason this method works is that it keeps every step measurable and reversible. By automating one task, supervised, against a recorded baseline, you always know whether it is working and can pull back if it is not. Owners who try to automate everything at once lose that clarity, because when several things change together you cannot tell what helped or what to fix. Sequential, measured automation is slower to start but far faster to a result you can trust, and each completed task returns the time and confidence that fund the next.
Implementing AI in a small business is a focus problem, not a technology problem. Pick one repetitive task, give it to one tool, measure the hours it returns, and only then expand. Start with the bottleneck, not the platform, and the project pays for itself before you have finished setting it up. At AS Consulting we build these implementations for UK small businesses one task at a time, for exactly that reason: the disciplined, boring approach is the one that works.
If you want this mapped to your own business rather than worked out from scratch, that is exactly what AS Consulting does, one task at a time. Pick today's single most repetitive job, write down what it currently costs you in hours, and you have already taken the first and hardest step.
Full article on LinkedIn · Video on YouTube · More from AS Consulting.
Automate smarter. By Simon Weiner, founder of AS Consulting (asconsulting.top), London.