How to Choose AI Automation for a Small Business: The Plain-English Checklist

Most small business owners I talk to know AI automation could save them time. They have read the posts, watched the videos, sat through the LinkedIn carousels. They are still standing at the same fork they were six months ago. Should they automate the receptionist, the leads, the quotes, the invoices, the diary, all of it, none of it? And which agency, tool, or platform?

This is the checklist I wish someone had given me when I started looking at AI for our own building firm. No vendor jargon. No "synergy". Just the questions that actually decide whether automation fits, what to build first, and who to hire.

Step 1: Find your bottleneck before you find a tool

The biggest mistake is starting with a tool. "We need a chatbot" or "we should get an AI receptionist" is the answer to a question you have not asked yet. The right starting point is the bottleneck.

Sit down with a notepad. List every recurring task in your week that costs you more than an hour or makes you sigh when you see it on the list. Beside each one, write a rough hourly cost (your time at £30 to £80 an hour, plus any team time involved). Add the revenue impact: does this task win jobs, lose jobs, or sit neutral?

Sort the list by hours times revenue impact. The top three are your candidates. The tools come after.

Step 2: Apply the readiness test

You are ready to automate if you can tick all four of these.

• Repeating revenue. More than ten paying jobs a month. Less than that and the time you save will not pay for the build.
• Known pain points. You can name two recurring tasks that frustrate you. If you cannot, you are not ready, you are bored. Different problem.
• Basic digital infrastructure. A phone, email, calendar. Ideally a CRM. If your bookings live in your head, automate the bookings system first, then come back.
• Decision authority. You can sign off changes without asking three other people. Decision-by-committee is the slowest part of any automation project.

Miss any one and the build will struggle. Three or fewer ticks means do the foundation work first.

Step 3: Match the bottleneck to the right type of automation

Most small-business automation falls into one of five buckets. Match yours.

Numbers are UK-typical for a one to fifteen person business. Bigger team or unusual industry, expect higher.

Step 4: Test the agency before you buy

Some questions to ask any AI agency before you sign anything. The answers tell you more than the slides will.

1. Who owns the build after delivery? If the answer is not "you do", walk. Lock-in dressed up as a service.
2. What happens if I cancel? Real answer: you keep what we built, we hand over the credentials, you can hire anyone. Anything less is a leash.
3. How do you handle UK GDPR? If they say "what is GDPR", run. If they say "we have a DPA template and we sign it before we touch your data", that is the right answer.
4. What is the rollback plan if something breaks? A real agency has tested this. They can name the rollback steps without thinking.
5. Can I see a working example with a real client? Demos in their own dashboard are fine. But ask whether you can see a live customer install. If they say "confidential", ask for an anonymised case study with screenshots.
6. How do you decide what to build first? If they recommend the same thing to every client, that is sales, not consulting. The right answer involves your numbers.

Step 5: Decide build vs. buy vs. hire

For each of your top three bottlenecks, the decision tree is short.

• Build yourself if it is one tool, you have the time, and the stakes are low. Pulling a Calendly + Zapier together for a booking flow is a Saturday morning job.
• Buy off the shelf if a mature SaaS already does what you need (quoting, invoicing, basic CRM). No need to build what £30 a month already does.
• Hire an agency if it is more than one tool, the stakes touch revenue, or your time is better spent on customers. The maths usually wins for the agency on anything above £2,000 of build work.

Step 6: Set the success metric before you start

Write down the number you expect to move and by when. Without this, you will never know if the build worked. Sample shapes that work:

• "Lead response time falls from 4 hours to under 1 minute on 90% of leads, within 30 days of go-live."
• "Missed-call rate falls from 28% to under 5%, within 60 days."
• "Time spent on invoice chasing falls from 6 hours a week to under 1 hour, within 14 days."

If your agency cannot put a number on the outcome before they start, they are not measuring. Either you give them the number or you walk.

Step 7: Start small, prove it, expand

The biggest cause of failed AI automation projects is scope. "Let's automate everything" projects almost always stall. "Let's fix the worst bottleneck and prove it pays back, then move to the next" projects almost always succeed.

Start with one. Track the metric. If the number moves, build the next one. If it does not, find out why before doing anything else. The Information Commissioner's Office publishes useful guidance on small-business AI use under UK GDPR (see the ICO guidance for organisations) and the British Standards Institute has BS ISO/IEC 42001 on AI management systems. Both are worth a skim before any install.

The honest verdict

The right first AI automation for your small business is the one that pays back fastest with the least risk to your existing operation. For most UK trades, that is Speed to Lead or AI Voice Receptionist. For most small businesses, it is the lead response or the customer comms. The Opportunity Map exists to remove the guesswork and give you the answer in 45 minutes.

Anyone selling you "the future of AI" is selling. Anyone helping you fix the bottleneck that is costing you money this week is helping. Pick the helper.