Small business automation: what to automate first, and what to leave alone
Small business automation works when you start with one job, not one tool. Score your repetitive jobs on how often they happen, how long they take, how often they go wrong and how risky they are. Automate the highest scorer first, keep a person checking anything risky, and time the job before and after.
I'm Managing Partner of a two-office law firm in East Sussex with around 30 people. I built the automations it runs on: a live reporting dashboard fed from our case management system, task hand-offs with one-click sign-off, and an AI-assisted marketing workflow with automatic checks.
Sound familiar?
You know you should be automating something. Every tool promises to do everything, and the last software you bought is still waiting for someone to use it properly.
- There are hundreds of apps, and you can't tell which ones matter for you.
- You paid for a system once, and the team drifted back to spreadsheets.
- You're not technical, and you don't want to become technical.
- You worry that an automation will go wrong in front of a customer.
The answer isn't a cleverer tool. It's a better way of choosing the job.
Is automation for small business worth it yet?
It can be, as long as you start with a real job rather than a tool. The ONS found that in June 2026, 28% of UK businesses with 0 to 9 employees used at least one AI technology. But plenty of useful automation needs no AI at all: reminders, and information moving between systems.
That compares with 49% of businesses with 250 or more employees, in the same survey. Both figures are from the Office for National Statistics article Artificial intelligence in UK businesses: 2023 to 2026, published in July 2026.
The ONS also lists the barriers businesses report most often: difficulty identifying where AI would help, cost, and a lack of expertise. The first of those is what the rest of this guide is for.
How do you decide what to automate first?
Score each repetitive job from 1 to 3 on four things: how often it happens, how long it takes, how often it goes wrong and how risky it would be if an automation got it wrong. Add the first three, subtract the risk, and start with the highest total. Risky jobs keep a person approving.
First, list every job that happens the same way each time. Ask the people who do them, because they know the workarounds you don't. Then score each one.
| Question | Score 1 | Score 2 | Score 3 |
|---|---|---|---|
| How often does it happen? | Monthly or less | Weekly | Daily or more |
| How long does it take each time? | Under 5 minutes | 5 to 30 minutes | Over 30 minutes |
| How error-prone is it? | Rarely goes wrong | Sometimes missed or late | Often retyped, missed or late |
| How risky if an automation got it wrong? | Low: easy to spot and undo | Medium: a customer might notice | High: money, personal data or a decision about a person |
Priority = how often + how long + how error-prone minus how risky. A total of 7 or 8 means start here, 5 or 6 is next in line, and 4 or under can wait. Any job that scores 3 for risk keeps a person approving each item, whatever its total.
| Job | Often | Long | Errors | Risk | Priority |
|---|---|---|---|---|---|
| Booking confirmations and reminders | 3 | 2 | 3 | 1 | 7: start here |
| Logging website enquiries | 3 | 1 | 3 | 1 | 6: next |
| Chasing overdue invoices | 2 | 2 | 3 | 1 | 6: next |
| Month-end management report | 1 | 3 | 2 | 1 | 5: next |
| Paying supplier invoices | 2 | 2 | 2 | 3 | 3: a person approves |
| Annual insurance renewal | 1 | 2 | 1 | 2 | 2: leave alone |
Look at the supplier payments. They take real time, but the risk score stops them jumping the queue. You can automate the paperwork around them, while a person still approves every payment.
Not sure how your own jobs would score? We can score them together on a free call.
Book a free 30-minute callWhich business automation ideas suit a small business?
The ideas that usually fit are the dull, frequent ones: logging enquiries, booking reminders, invoice chasing, bank reconciliation, filing, approvals and onboarding. Many run on software small firms often already have: Microsoft 365, Google Workspace, Xero or QuickBooks. Here are 18, grouped by area, to score against your own list.
Sales
| Idea | What happens | Often done with |
|---|---|---|
| 1. Log every enquiry | Each web form or email becomes a record in your CRM, and the right person gets a nudge. | Zapier, Make or Power Automate |
| 2. Draft the first reply | AI drafts a reply from your own notes. A person checks it and sends it. | ChatGPT, Microsoft 365 Copilot or Claude |
| 3. Follow up quiet quotes | A quote with no answer after a week creates a follow-up task. | Your CRM, or Zapier or Make |
| 4. Log and thank referrals | Every referral is recorded, the referrer is thanked, and a follow-up date is set. | A CRM with email automation |
Zapier describes each automation as a trigger followed by actions. That's a handy way to write any of these down: “when a form arrives, add a row and email the sales inbox”. I've been in BNI for 20 years, so if referrals bring you work, I'd score idea 4 early.
Customer service
| Idea | What happens | Often done with |
|---|---|---|
| 5. Online booking with reminders | Customers pick a slot, and confirmations and reminders go out by themselves. | Microsoft Bookings, included in Microsoft 365 Business Basic, Standard and Premium |
| 6. Saved answers to common questions | The replies you type every week, inserted in a click and edited as needed. | Gmail templates |
| 7. Sort incoming email | Messages from key customers, or with certain words, go straight to the right folder. | Outlook rules |
| 8. Ask every customer for a review | A review request goes after every finished job, not just the happy ones. | Your job system or CRM, plus email |
On reviews, the rules tightened in April 2025: fake reviews, and reviews that hide an incentive, are now banned practices. The CMA's guidance on reviews and endorsements explains what businesses must do.
Finance
| Idea | What happens | Often done with |
|---|---|---|
| 9. Invoice reminders | Polite reminders before and after the due date, which stop once the invoice is paid. | Xero (up to five reminders) or QuickBooks (up to three) |
| 10. Repeating invoices | Customers on a regular arrangement get the same invoice each month without anyone raising it. | Xero repeating invoices |
| 11. Bank feeds and rules | Transactions arrive in your accounts, and rules code the routine ones the same way each time. | Xero bank feeds and bank rules |
| 12. Digital receipts and bills | Receipts and bills go straight into your accounts software instead of a drawer. | Your accounts software, or an app your accountant recommends |
Idea 12 matters more now. Sole traders and landlords with qualifying income over £50,000 have had to use Making Tax Digital for Income Tax from 6 April 2026, keeping digital records in compatible software. The threshold falls to £30,000 from April 2027 and £20,000 from April 2028. For idea 9, my guide to chasing unpaid invoices has a reminder schedule you can copy.
Admin
| Idea | What happens | Often done with |
|---|---|---|
| 13. File attachments automatically | Signed forms, supplier invoices and photos from email are saved to the right folder. | Power Automate or Zapier, with OneDrive, SharePoint or Google Drive |
| 14. Turn forms into documents | A form fills in a job sheet, letter or checklist from your template. | Microsoft Forms or Google Forms, with Power Automate or Zapier |
| 15. Weekly figures without spreadsheets | The numbers you check every Monday are pulled together for you. | A simple dashboard linked to your systems |
Team
| Idea | What happens | Often done with |
|---|---|---|
| 16. One-click approvals | Holiday, expenses and sign-offs go to the right person, who approves from Outlook or Teams. | Power Automate approvals |
| 17. New starter checklist | Each person gets their onboarding tasks on the right day, with a reminder if one slips. | A shared checklist with automatic reminders |
| 18. Meeting notes and actions | AI drafts the notes and actions. The person who chaired the meeting checks them before they go round. | An AI notes tool in Teams or Zoom |
Ideas 15 and 16 are close to what I built in my own firm: a dashboard that replaced hand-built reports, and hand-offs that replaced emails asking for sign-off. If you use Microsoft 365, my Power Automate examples show more of this in practice.
What should you not automate?
Leave alone anything that needs judgement, empathy or a real conversation, anything that happens only once or twice a year, and any process that's already a mess. Take extra care with money and people: payments to new bank details, and decisions about individuals such as hiring, should always have a person making the call.
- Decisions about people. Under UK data protection law, decisions made solely by automation that have a legal or similarly significant effect on someone carry extra safeguards, including the right to human intervention. The ICO's guidance on automated decision-making sets them out.
- Paying new or changed bank details. The National Crime Agency advises checking changed payment details by calling the supplier on a number you've used before. Automate the paperwork, never that check.
- Marketing emails without consent. Automation doesn't change the rules. Individuals and sole traders need to have agreed, or be existing customers under the soft opt-in.
- Complaints, bad news and disputes. Use automation to alert the right person quickly, not to reply for them.
- Rare jobs. If it happens twice a year, a written checklist is simpler than an automation.
- A broken process. Automating a muddle gives you a faster muddle. Fix the steps first.
AI drafts, people decide. Acas advises employers to check AI output for accuracy, bias and tone. Anything AI writes for a customer should pass a person first. This is general information, not legal advice.
How do you measure the time automation saves?
Time the job before and after. For two normal weeks, count how many times it happens and how long each takes. Once the automation has settled in, measure two similar weeks again, including the time spent checking its work and fixing hiccups. Then subtract the time the automation itself needs to keep running.
| Measure | Before | After |
|---|---|---|
| How many times the job happened | Count for two normal weeks | Count for two similar weeks |
| Minutes per job | Timed from start to finish | Timed, including checking the automation's work |
| Mistakes or misses | Count them | Count them |
| Upkeep | None | Minutes spent fixing, updating or answering alerts |
Monthly time saved = (minutes per job before minus minutes per job after, including checking) × jobs per month, minus upkeep minutes per month.
Three honesty rules help. Compare like with like, so a quiet fortnight isn't measured against a busy one. Don't count time that has only moved to someone else's desk. And count fewer mistakes as a win in its own right, even when the minutes barely change.
In practice: what I see
In my own firm, reporting was an easy place to start. It touched everyone and the job was the same every week, so it would score well on the table above.
Not every idea survives contact with a busy team. If people quietly work around a new system, it has added a step instead of removing one. Fix it or drop it quickly rather than insisting.
Before anything gets built, I'd ask four questions about it:
- Who owns it? One named person who notices if it stops.
- What happens when it fails? An alert, and a manual fallback everyone knows.
- Where does a person check? Before anything goes to a customer, or moves money.
- How do we switch it off? Written down in plain English.
Common pitfalls
- Buying the tool first. Pick the job, then the tool. Often the tool is one you already have.
- Automating everything at once. Build one, test it on real work with the people who do it, then add the next.
- Silent failures. If an automation breaks quietly, you'll find out from a customer.
- Personal accounts. Build on business accounts, with UK GDPR in mind, and follow the ICO's guidance on AI and data protection where AI is involved.
- Leaving the team out. Acas advises consulting staff before introducing AI, and the same courtesy helps any automation stick.
What to do next
How I can help
If you'd rather not score it alone, my AI audit, a readiness assessment for small businesses, does it with you. I map how your week goes, score the jobs and hand you a plan with the quick wins first.
When you're ready to build, business automation is where I connect the systems you already use so the work moves by itself, with a person checking anything that matters.
Questions people ask
Do I need to be technical to automate my business?
No. Many automations are a setting you switch on in software you already use, such as invoice reminders in Xero or booking reminders in Microsoft Bookings. For anything linking several systems, you need someone who can build and document it, and a person on your side who owns it.
How is automation different from AI?
Automation is a set of rules: when this happens, do that, every time. AI is useful where the input is messy, like an email in someone's own words, and it gives you a draft rather than a certainty. Sensible set-ups combine the two and keep a person checking the drafts.
Which automation tool should a small business use?
Use what you've got first. Businesses on Microsoft 365 often begin with Power Automate, while Zapier and Make suit a mix of separate online apps. My comparison of Zapier and Make goes into more detail.
Will automation replace my staff?
That shouldn't be the aim. The ONS found improving business operations was the most common purpose businesses gave for using AI. The people doing the work know where the time goes, so involve them from the start.
How much does small business automation cost?
It depends on the job and the tools. Many features are included in software you may already pay for, so check that first. Price anything new from the supplier's own website before you commit. My own work is quoted in writing after a free call.
Related guides
12 Power Automate examples for small businesses on Microsoft 365
12 Power Automate examples for a small business on Microsoft 365, each set out as trigger and actions in plain English, with the licence each one needs.
Read the guideMoney and admin · 12 min readChasing unpaid invoices: late payment interest and automatic reminders
Chasing unpaid invoices from other businesses? How statutory late payment interest works, a worked example and a polite reminder routine for Xero or QuickBooks.
Read the guideAI readiness assessment
An AI readiness assessment for small businesses in Sussex and Kent. Find where your week goes, what AI can safely take on, and get a clear 90-day plan.
AI readiness assessment