What to automate first is the question every owner reaches once the admin starts eating their evenings. Almost every article on the subject is really a list of fifteen tools. It is the wrong shape of advice, because the tool is the last decision, not the first, and buying software before you have identified the task is how businesses end up paying three subscriptions for things nobody uses.
Here is a different way to decide what to automate first, and some UK numbers worth knowing before you start. It works for a two-person joinery firm in Leicester as well as a thirty-person wholesaler in Leeds, because it starts with your week rather than someone's software catalogue.
On this page
- Where UK small business automation stands
- Find the task before you find the tool
- What to automate first: the five that usually come out on top
- What not to automate yet
- The UK data protection part, briefly
- What to automate first, by type of business
- How to score your list properly
- Tools, once you know the task
- A worked example from a typical trade business
- Where AI automation helps and where plain rules do the job
- A 30 day plan for your first automation
- Mistakes we see when owners pick what to automate first
- Questions to ask before you pay anyone for AI automation
- What to automate first when you work alone
- Who should own each automation
- Signs an automation isn't working
- Start with one thing, and measure it
- Related reading
- Not sure what to automate first?
Where UK small business automation stands
The government's SME Digital Adoption Taskforce reports that 43% of SMEs surveyed by the British Chambers of Commerce in 2024 had no plans to use AI at all, down from 48% the year before. Movement, but slow movement. It also means most of your competitors have not started either.
The same report puts the prize in context. Raising SME productivity by one per cent over five years would add £94 billion a year to UK GDP, and the research it cites finds firm-level productivity improvements of 7 to 18 per cent per technology adopted, depending on the product. Federation of Small Businesses research found SMEs that innovate recorded 14.8% revenue growth.
Meanwhile the UK ranks 25th worldwide for future digital readiness in the IMD 2024 index, and Be the Business found UK SMEs invest less in technology and management than their G7 peers.
Read together: the gains from small business automation are real and measurable, most of your competitors have not moved, and the main barrier the taskforce identifies is not cost but the sense that starting is too difficult and too risky.
Find the task before you find the tool
For one week, write down every task that meets three conditions: you do it more than twice a week, it follows the same steps every time, and it produces a predictable output. Do not filter, just log. The log is the only honest way to see where your time goes.
At the end of the week you will have somewhere between ten and thirty entries. Multiply each by how long it takes and how often it happens. The top three are your answer to what to automate first. Everything below them is a distraction until those three are done, however tempting it looks.
This matters because the tasks that feel most annoying are rarely the ones costing the most time. Chasing an invoice feels worse than retyping delivery addresses, but the retyping is probably eating four times the hours.
What to automate first: the five that usually come out on top
Across the UK businesses we have worked with, the same handful keep appearing.
Retyping data between systems. An enquiry arrives by email, gets typed into a spreadsheet, then typed again into the quoting system, then again into the accounts package. Every retype is a chance to introduce an error and a job nobody wants. This is not really AI, it is plumbing, and it is usually the highest return thing on the list. For most firms, this is what to automate first.
First response to enquiries. Not replacing you, acknowledging receipt. An immediate reply that confirms the message arrived and says when a human will respond keeps you in the running against competitors who answer within the hour. Layer in answers to the five questions you get asked constantly and a meaningful share never needs your attention.
Quote and invoice preparation. If your quotes are assembled from the same components with different quantities, that is a template problem, and templates are cheap to build.
Chasing payment. A polite reminder at 7, 14 and 21 days past due, sent automatically, recovers money that otherwise sits unclaimed because nobody enjoyed making the call. Under UK late payment legislation you are entitled to statutory interest and a fixed recovery charge on overdue commercial invoices, and simply stating that in the third reminder tends to speed things up.
Turning notes into documents. Site visit notes into a report, a phone call into a summary, meeting notes into actions. This is where the current generation of AI tools earns its keep, because it is language work with a human checking the output.
What not to automate yet
Knowing what to automate first also means knowing what to leave alone: anything where being wrong is expensive and hard to spot. Pricing decisions, anything that goes to a customer without a human reading it, anything involving personal data you have not thought through, and anything where the process itself is still changing month to month. Automating an unstable process just makes the instability faster.
Also be careful about publishing AI-written content unreviewed. It reads as generic, which costs you the credibility that makes small businesses worth choosing in the first place.
The UK data protection part, briefly
If customer data goes into a tool, UK GDPR still applies. Check where the provider processes and stores data, whether it trains models on what you submit, and whether you need to update your privacy notice.
The ICO publishes guidance aimed at small organisations that is clear and readable. If you also sell into the EU, the EU GDPR applies to those customers too, and the checks are much the same. This is not a reason to avoid automation, but it is a reason to check before you connect your customer list to something.
None of this is legal advice, and if you handle sensitive data it is worth a short conversation with someone who does that professionally.
What to automate first, by type of business
The log will give you your own answer, but the patterns by sector are consistent enough to be a useful starting guess.
Trades and field services
Plumbers, electricians, gardeners and cleaning firms usually lose most time to enquiry handling and job paperwork. What to automate first here is almost always getting enquiries into one place and turning job notes into invoices without retyping.
Shops and online stores
For a small retailer in Norwich or a WooCommerce store selling homeware, stock updates and order emails are usually what to automate first. If you sell on more than one channel, keeping stock levels in sync saves both time and overselling.
Professional services
Accountants, consultants and agencies tend to bleed time on onboarding and chasing documents from clients. Reminders and a standard intake form are what to automate first, well before anything involving AI.
How to score your list properly
Hours saved is the obvious measure, but it's not the only one. Before you decide what to automate first, give each task on your list a quick score out of three on four things. It takes ten minutes and stops you picking the task that's merely the most irritating.
Time
How many hours a month does it eat? Be honest and use the log, not your memory. Memory always overestimates the annoying jobs and underestimates the boring ones.
Cost of mistakes
What happens when someone gets it wrong? A typo in a delivery address means a failed delivery and a redelivery fee. A typo in an invoice total means an awkward phone call. Tasks where errors are common and expensive score higher.
Stability
Has the process stayed the same for the last six months? If yes, score it high. If it changes every time a new supplier comes on board, score it low, because you'll be rebuilding the automation every quarter.
Who does it now
If the task sits with the owner, or with your most skilled person, it scores higher. An hour of the owner's time freed up is worth more than an hour of a junior's, because of what the owner does with it.
Add the scores up. The top one is what to automate first. It's often not the task you'd have guessed, and that's the point of doing it on paper.
Tools, once you know the task
Only once you know what to automate first does the tool question come in. For most small businesses it's one of three routes.
Start with the first route, because your first automation should never need a new subscription to prove itself. Move to the second when you hit a gap. Only move to the third when the first two can't do it, or when the monthly bills for connector tools start to look silly.
- What you already pay for. Xero, QuickBooks, Microsoft 365 and Google Workspace all have automation built in that most owners never switch on. Invoice reminders in Xero take ten minutes to set up.
- Connector tools. Zapier, Make and n8n move data between apps when something happens: a form is filled, an email arrives, a payment lands. They're good for the retyping problem and cheap to try.
- A small custom build. When the task touches your own systems, or needs rules your apps can't express, a short piece of custom software is often cheaper over three years than stacking subscriptions.
A worked example from a typical trade business
Picture a kitchen fitter in Leicester with four staff. Enquiries come in through the website form, Facebook and the phone. The owner types each one into a spreadsheet, then into the quoting template, then into Xero once the job's booked.
Run the one-week log and the retyping comes out top: three systems, every enquiry, with errors creeping in on measurements and postcodes. Chasing payment comes second. Writing up site visit notes comes third.
So what to automate first is clear. Web form and Facebook enquiries go straight into one sheet, and booked jobs push into Xero automatically. Invoice reminders get switched on in Xero. The site notes wait until the first two have run cleanly for a month.
That's not a big project. It's two or three small ones, done in order, each one checked before the next starts.
Where AI automation helps and where plain rules do the job
Not every task needs AI. A lot of what gets sold as AI automation is simple rules: when a form arrives, copy these fields there. That's fine, and it's cheaper and more predictable than a language model.
Use rules when the input is tidy
If the data arrives in the same shape every time, such as a web form, a payment notification or a CSV from a supplier, rules win. They don't guess and they don't drift. Most of the retyping problem is solved this way.
Use AI when the input is messy
AI earns its place when the input is free text. Emails that ask five things in one paragraph, handwritten site notes typed up on a phone, a customer's rambling voicemail. A language model can pull out the useful bits and draft a reply or a report for a person to check.
The rule of thumb is simple. If you can write the steps down as a flowchart, use rules. If you'd need to explain it to a new starter with examples, AI might help. Either way, decide what to automate first before you pick the kind of automation.
A 30 day plan for your first automation
This is the plan we'd give any owner who asked us what to automate first and how to go about it.
At the end of the month you'll know whether it worked, and you'll have a method for the next one. That's worth more than the time saved on the first task.
- Week one. Keep the task log. Every repeated task, how long it took, how often.
- Week two. Score the list and pick one task. Time it properly across the week so you've got a before figure.
- Week three. Build the simplest version that could work, using tools you already pay for if you can. Run it alongside the manual process.
- Week four. Stop doing it by hand. Check the output every day. Time how long the remaining work takes.
Mistakes we see when owners pick what to automate first
The fix for all five is the same: log, score, pick one, measure. It's dull, and it works.
- Starting with the most exciting task instead of the most expensive one.
- Automating a process only one person understands, without writing it down first.
- Buying an annual subscription before a free trial has proved anything.
- Trying to automate customer conversations before the back office is sorted.
- Measuring success by how clever the setup looks rather than hours saved.
Questions to ask before you pay anyone for AI automation
If you bring someone in to help, these questions sort the useful from the expensive very quickly.
A good answer to the first question should come from your task log, not a product demo. If they can't tell you what to automate first without selling you a platform, keep looking.
- What would you automate first in my business, and why that one?
- How will we measure whether it worked?
- What happens when one of the connected apps changes or goes down?
- Where does my customer data go, and is it used to train anything?
- Can my team change simple settings without calling you?
What to automate first when you work alone
Sole traders have it harder. There's no admin person to hand things to, and every hour on paperwork is an hour not billed. A freelance bookkeeper in Bristol or a mobile hairdresser in Cork can't log tasks for a week and then book a planning meeting. So the question of what to automate first gets squeezed into the evenings.
Keep the log anyway, just shorter. Three days is usually enough to show what to automate first when you do everything yourself, because the same jobs repeat quickly. Look hard at anything you do on your phone between appointments. That's often retyping in disguise.
For most people working alone, what to automate first is either the booking confirmation or the invoice. Both are tidy, both repeat, and both are already built into tools like Xero, Square or a decent booking app. Switch the feature on before you look at anything new.
Don't build a system for a team you haven't hired. Solve this month's problem. When you do take someone on, revisit what to automate first, because the answer changes once another person shares the work.
Who should own each automation
Every automation needs a named person who notices when it breaks. Not the developer who built it. Someone inside the business who sees the output every day.
Automations fail quietly. A password changes, an app updates its form, a supplier sends a different file format, and the flow stops. Nobody notices for a fortnight because the whole point was that nobody had to look. Give one person a weekly two-minute check: did the expected things happen? If they didn't, they know who to call.
Signs an automation isn't working
Any of those means stop and rethink. A bad automation is worse than none, because it looks finished. Get what to automate first wrong and you pay for it twice.
- Staff are still doing the task by hand "just to be safe".
- Someone spends time every week fixing what the automation produced.
- Nobody can explain what it does or where the settings live.
- It saves time for one person and creates work for another.
Start with one thing, and measure it
The single task at the top of your list is what to automate first. Automate that, and nothing else for now. Time how long it takes now, across a fortnight. Automate it. Time it again.
If it did not save meaningful time, you picked the wrong task. Stop and try the next one rather than adding a second tool on top of a first that is not working. Businesses that get value from small business automation almost always got it from two or three things done properly, not from a stack of subscriptions.
Related reading
- What the tasks above look like once they are built with language models rather than rules: LLM agents and where they cut operational cost.
- Automating the first response only pays if enquiries are arriving at all: five things we check when a site has traffic but no enquiries.
- The retyping problem in a trading business usually traces back to the accounting package: Busy, Tally, QuickBooks or custom.
Not sure what to automate first?
Tell us how your week runs and we will tell you what to automate first, which parts are worth automating later and which are not. If the honest answer is that your processes are not stable enough yet, we will say so rather than sell you something. Free consultation, no commitment.
Book a free consultation at devsioservices.com/contact, read about our AI automation for small businesses, or see when a custom software build makes more sense. Fixed-price options are on the pricing page.
Related guides
- How to automate lead intake and invoicing with AI. A practical four-step pipeline.
- Custom software development for small businesses in the USA. When it pays off and how to buy it.
