OptimisedThe owner of this business was doing his invoicing at half past nine on a Tuesday night when he first emailed me. He'd been running a heating and plumbing firm in the Midlands for eleven years, he had two staff, and by his own reckoning he hadn't taken a full weekend off since 2023.
Eight months later, the three of them are doing roughly 20 fewer hours of admin a week between them. Not because they hired anyone, and not because we built anything clever. Most of what we did was switch on features they were already paying for, connect four tools that cost less than £75 a month combined, and stop them doing three things they didn't need to do at all.
I want to walk through exactly what happened, including the two things that didn't work, because I think this is a far more useful small business automation case study than the glossy ones you'll find on vendor websites. The names and a few identifying details are changed with the owner's permission. The numbers are real.
The Short Version
They went from about 26 hours of admin a week across three people to about six. The biggest single win was invoice chasing, which took nearly seven hours a week and now takes almost none. The rest came from booking, quoting, customer emails and moving information between systems. Total monthly spend on tools is around £70 plus VAT, with a small per payment fee on Direct Debits. The setup took eight weeks of part time effort, most of it in the first three.
If you have fewer than five staff and you're spending evenings on paperwork, the first two changes alone will probably get you more than half of this.
Who They Are And Where They Started
I'll call the owner Martin. He's a Gas Safe registered engineer in his early fifties, a sole trader with turnover comfortably above £50,000, which matters later because it puts him inside Making Tax Digital for Income Tax. He employs one full time engineer, and his wife does the office side three days a week.
Before we changed anything, I asked all three of them to keep a simple log for a fortnight. Just a notebook and a rule: every time you do something that isn't a job, write down what it was and roughly how long it took. Most people skip this step, and it's the one that matters most, because you can't tell whether automation worked if you never measured what it replaced.
Their fortnight came out at an average of 26 hours a week: about 12 for Martin, five for the engineer and nine for the office manager. That's higher than the national picture but not wildly so. According to the SME Business Barometer from American Express and Small Business Saturday UK, which surveyed 1,000 UK owners in 2026, the average is 11 hours a week on admin and finance tasks, roughly twice what those owners spend on winning new work. A separate survey by HeyBRB of 167 UK owners, weighted towards trades, found eight hours a week going on repetitive admin that owners already knew could be handled differently. Three people, three people's worth of admin.
Here's how the 26 hours broke down, in order of size:
Chasing unpaid invoices and reconciling payments, close to seven hours. Answering the phone and email to book jobs, then juggling the diary, about five. Writing quotes, three and a half. Answering the same customer questions over and over, about three. Re typing job details from the engineers' paper sheets into invoices, about three. Receipts, bookkeeping and the general dread of the tax return, around two and a half. Everything else, a couple of hours.
The invoicing figure surprised Martin. He knew it was a problem but had no idea it was the biggest one. The HeyBRB report found that 83 percent of owners have never calculated what their admin time costs them, which matches what I see. People know it hurts. They rarely know where.
Why Invoicing Came First
I always start where the hours are largest and the judgement required is smallest. There's no skill in sending a polite reminder seven days after a due date, and no relationship value in it either. It's just a job that has to happen, and it was the job Martin's wife hated most.
They were already on Xero. What they weren't doing was using it properly. Their invoices were going out by email with no payment link, reminders were being typed by hand when someone remembered, and payments were being matched to invoices manually once a week from a bank statement.
We did three things. First, we switched on Xero's automatic invoice reminders, set to fire at seven days overdue, fourteen days overdue, and then a firmer one at twenty one days that also copies Martin so he can decide whether to pick up the phone. Second, we added an online payment option to every invoice so a customer can pay by card from their phone the moment the invoice arrives. Third, and this was the big one, we moved their boiler service plan customers onto Direct Debit through GoCardless.
Martin has about 140 households on an annual service plan paying monthly. Every one of those had been a standing order set up by the customer, which meant every one of them broke when a customer changed bank, moved house or simply forgot. GoCardless charges 1 percent plus 20p per UK Direct Debit, capped at £4, which on a typical £15 monthly plan payment works out at 35p. In exchange, the collections happen automatically, the mandate follows the customer, and the payment lands in Xero already matched to the right invoice.
Time saved on this cluster: about six hours a week. Cash flow improved too, which is the benefit owners actually feel first. Within two months the time to get paid on one off jobs had dropped noticeably, though I'm wary of quoting a precise figure because the sample was small and winter is always their busiest period.
The cost of the Xero side was zero beyond what they already paid. They're on the Grow plan, which Xero's UK pricing page lists at £39 a month excluding VAT following the September 2026 price rise. Grow includes automatic reminders, unlimited invoices, bank feeds, Hubdoc for receipt capture and payroll for one person. Martin had been paying for all of that for four years and using about a third of it.
Booking Without Answering The Phone
The second largest drain was the diary. Someone rings about a boiler service, Martin's wife checks the engineers' availability on a whiteboard, calls back, the customer's changed their mind about the day, and so on. Multiply that by fifteen or twenty bookings a week and you have five hours gone.
This is where a lot of automation advice goes wrong for trades, so let me be clear about what we didn't do. We did not put in an AI phone answering system. Martin's customers are largely older homeowners who want to talk to a person about the strange noise their boiler is making. A robot on the phone is a fast way to lose them.
What we did instead was give customers a way to book routine work themselves, while keeping the phone exactly as it was for anything else. Annual services, landlord gas safety checks and radiator bleeds are predictable jobs of predictable length. For those, we set up a Calendly booking page linked to the engineers' calendars, with a two hour slot per service and buffer time built in for travel. The link goes in the annual service reminder email, on the website, and in the automatic text that goes out when a service falls due.
Calendly's Standard plan is $10 per seat a month billed annually, or $12 monthly, which is roughly £8 to £10. Calendly bills in US dollars, so the pound figure moves with the exchange rate, and the same is true of the other dollar priced tools I'll mention. They need one seat because the two engineers share a single "next available" calendar rather than each having their own.
About 60 percent of routine bookings now come through the link. The phone still rings for breakdowns and for people who prefer it, and that's fine. That was never the target. Time saved: a little under four hours a week, and the whiteboard has been retired.
Quotes That Take Twenty Minutes Instead Of An Hour
Martin's quotes were taking him an hour each, typically written on the sofa after tea. Part of that was the pricing, which needs his judgement and always will. Part of it was the descriptive text: what the work involves, why he's recommending it, what's included, what the customer needs to do beforehand. That text was nearly identical from quote to quote, and he was writing it fresh every time.
We built him six quote templates in Xero for his most common jobs, each with the standard line items already in place and a description block that only needs the specifics changed. For the less common jobs, he now dictates rough notes on his phone in the van and uses ChatGPT Plus to turn them into a clean, plain English description that he then checks and pastes into the template. ChatGPT Plus is $20 a month, around £16 before VAT, and again it's billed in dollars.
Two warnings. AI drafts the words, not the price; I've seen people try to get a chatbot to price plumbing work and the results are dangerous. And Martin reads every word before it goes out, because the tool will occasionally invent a detail, add a promise he didn't make, or name the wrong part. A quote is a contract. Ten minutes of checking is not optional.
Quotes now take him about twenty minutes, so a little under two and a half hours a week saved. He also sends them the same day rather than three days later, and he's convinced his conversion rate has improved. I believe him, but I'd want another six months of data before putting a number on it.
The Same Twenty Questions
Every business has a set of questions it answers constantly. For Martin it was things like whether he covers a particular postcode, what a service costs, whether he does landlord certificates, how long a boiler swap takes, and whether he's Gas Safe registered. About three hours a week was going on typing the same answers into emails and texts.
The fix was almost embarrassingly low tech. We wrote out the twenty most common questions with warm, accurate answers in Martin's own words. Those went onto a proper FAQ page on the website, into saved reply templates in their email, and into a short document the office manager keeps open. When a question comes in, she picks the template, adjusts one line if needed, and sends it.
We also tested an AI assistant that drafted replies to incoming emails automatically. It was decent on about 80 percent and confidently wrong on the rest, including one where it told a customer a job was under warranty when it wasn't. We turned it off. I'll come back to why.
Time saved anyway, purely from the templates: about two and a half hours a week.
Getting Information Off Paper
The engineers were filling in paper job sheets, which the office manager typed into Xero each week to raise invoices, and into a spreadsheet to track parts. Three hours a week, and a steady trickle of errors from handwriting and tiredness.
This is the one piece where we used what most people think of as "automation software". The engineers now fill in a simple form on their phones at the end of each job: customer, work done, parts used, time on site, photos. That form triggers a scenario in Make, which creates a draft invoice in Xero with the line items filled in, adds the parts to the tracking sheet, and sends the office manager a message saying the draft is ready to check.
I chose Make over Zapier for a boring reason: cost at their volume. Make's Core plan is $9 a month for 10,000 operations, roughly £7 before VAT. Zapier's Professional plan, which you need for anything more than a two step workflow, starts at $29.99 a month for 750 tasks, and a job sheet workflow with five or six steps eats that allowance fast. Both bill in dollars. Zapier is easier to learn and I'd still recommend it for someone building their own simple two step connection between two well known apps, but for a multi step workflow that someone else is setting up, Make is the better value by a wide margin. The SME Software Help comparison makes the same point with a worked example, and I've found their numbers match my own experience.
Two practical notes. Make's free plan gives you 1,000 operations a month and this workflow uses around 400, but I don't like a business's invoicing depending on a free tier that can change its limits at any time, so they're on Core. And the invoice is created as a draft, not sent. Automation should create the thing; a person should decide whether it leaves the building.
Time saved: about three hours a week, and the errors have all but disappeared.
The Tax Bit Nobody Wants To Talk About
Because Martin is a sole trader with qualifying income above £50,000, he's in the first wave of Making Tax Digital for Income Tax, mandatory from 6 April 2026, which means keeping digital records and sending HMRC a quarterly summary of income and expenses through compatible software, on top of the annual return. According to HMRC's own announcement on the first deadline, the first update covered 6 April to 5 July 2026 and was due by 7 August, and no penalty points are being issued for late quarterly updates in the first year. The threshold drops to £30,000 from April 2027 and £20,000 from April 2028, so if you're a sole trader below £50,000 now, this is coming for you too.
Martin had been dreading it. In practice, because his invoices, bills and bank transactions were all now flowing into Xero automatically, and receipts were being photographed into Hubdoc from the van, the quarterly update took his accountant about fifteen minutes. The digital record keeping that MTD demands is the same thing you do anyway when you automate your bookkeeping. It's one of the few cases where a compliance requirement and a productivity improvement point in exactly the same direction.
Bookkeeping and receipts went from about two and a half hours a week to under half an hour, most of it checking that Hubdoc has read the receipts correctly, which it mostly does and occasionally doesn't.
What It Added Up To
Here are the numbers side by side, because the point of a small business automation case study is that you can check the sums.
Invoicing and payments: from about seven hours to under one. Bookings: from about five to just over one. Quotes: from three and a half to about one. Customer questions: from three to half an hour. Job sheets and data entry: from three to almost none. Bookkeeping and receipts: from two and a half to half an hour. The remaining couple of hours of miscellaneous admin didn't change much, and I added roughly one hour a week for someone to check the automations are running and deal with the occasional exception. Call it 26 hours down to six.
The monthly running cost is Xero Grow at £39, Calendly at roughly £8, ChatGPT Plus at roughly £16, and Make Core at roughly £7, all before VAT, so about £70 a month, plus GoCardless fees that come to around £50 a month on their service plan volume. Against that, three people got back the equivalent of half a full time job between them, and Martin got his evenings back.
I should also be clear that the 20 hours didn't arrive on day one. After the first fortnight the saving was about nine hours. The templates, the booking link and the job sheet workflow each took a few weeks to bed in, partly because people needed to trust them before they stopped doing the old thing in parallel. By month three we were at 20, and it's held there since.
The Two Things That Didn't Work
I've mentioned both briefly, but they deserve their own section, because you'll be sold both of them.
The AI phone answering system was a clear no. We trialled one for a fortnight. It was impressive in a demo and irritating in real life, and two regular customers said so. Martin's phone manner is a large part of why people recommend him. Automating away the thing customers value most isn't efficiency. It's self sabotage with a subscription attached.
The AI email drafting was a softer no. It nearly worked, and with a member of staff whose whole job was checking and sending replies it would probably have earned its keep. In a business this size, the person checking the AI's work is the person who would have written the reply, so the saving evaporates and the risk of one bad answer remains. The right size of business for that tool is bigger than three.
What The Law Actually Asked Of Them
People ask me about GDPR the moment the word automation comes up, so here's what applied to Martin, and what didn't.
UK GDPR and the Data Protection Act 2018 are enforced by the Information Commissioner's Office, the ICO, and they apply to any processing of customers' personal data, automated or not. Martin was already a controller of that data before we touched anything. Connecting Xero to Make to Calendly doesn't change his obligations; it changes who his processors are, and each of those companies has a data processing agreement that he accepted when he signed up. The practical jobs were to check that his privacy notice on the website mentioned the tools he uses, and to make sure customer data wasn't being pasted into ChatGPT. The quote descriptions he generates contain a job description and an address at most, and we agreed he'd strip the address out before pasting. It's a small discipline but it matters, because putting personal data into a public AI tool is a disclosure to that provider and has to be treated as such.
The other change worth knowing about is that on 5 February 2026, section 80 of the Data (Use and Access) Act 2025 replaced Article 22 of the UK GDPR, which governed decisions made solely by automated means. The analysis from HLC explains that the new rules are more permissive about when businesses can use fully automated decisions with significant effects on people, provided they tell the person, allow them to challenge it and offer human review. None of this bites for Martin, because nothing in his setup makes a significant decision about a person without a human involved. But if you're thinking about letting software decline customers, set prices per person or screen applicants, this is the law you now need to read, and the ICO's AI guidance is being updated to reflect it.
How I'd Do It If It Were Your Business
A few questions I'd ask you to answer before you spend a penny.
Have you measured your admin for two weeks, honestly, in a notebook? If not, do that first. You'll be surprised, and the surprise will tell you where to start.
Is your largest time sink something that needs no judgement? Invoice chasing, appointment reminders, receipt capture and copying information from one place to another all qualify. Start there. Writing, pricing and talking to customers do not qualify, and I'd leave them alone for now.
Are you already paying for software you're not using? Log into your accounting package and look at the features list. Most of the businesses I work with are on a plan that already includes reminders, online payments, receipt scanning and templates. Switching those on costs nothing and often delivers a third of the total saving on its own.
Do you have someone who can spend an hour a week keeping an eye on things? Automations break. A bank feed disconnects, an app changes its login, a form gets edited and the workflow stops. Someone needs to notice, or the setup needs to be simple enough that failure is obvious.
Where I'd Push Back On The Usual Advice
Two opinions I hold that the automation industry mostly doesn't.
First, a three person business shouldn't buy an all in one platform. The marketing is seductive: invoicing, booking, email, chat, all in one place, one bill. In practice you get a mediocre version of six tools instead of a good version of the three you need, and you're locked in. Martin's four separate tools each do one thing well, and if Calendly disappears tomorrow he swaps in something else in an afternoon.
Second, hiring is still the right answer more often than my industry admits. If Martin's admin had been mostly judgement work, quotes, complaints, negotiations with suppliers, I'd have told him to take on a part time office manager and skip most of this. Automation removed the repetitive 20 hours. It didn't remove the six that are left, because those six are the ones that need a human, and they always will.
What To Do This Week
Start your notebook. Every admin task, every day, for a fortnight, with a rough time against each. Then open your accounting software and switch on automatic invoice reminders if they aren't already running. That's an hour of work and, if your business resembles the ones I see, it's the single highest return change you'll make this year.
If your biggest number turns out to be invoice chasing, bookings or data entry, you can copy this setup almost step for step. If it's writing, selling or talking to people, be careful what anyone tries to sell you, and consider whether the honest answer for a business your size is a person rather than a platform.