Business automation for UK small companies
Stop moving the same information between systems by hand
We find the repetitive work, connect the tools involved, and build controlled workflows that either finish the job or hand the awkward cases to a person. Nothing runs unsupervised that should not.
From £500plus VAT for a first build. Discovery £249, credited against builds over £1,000.
Manchester based, UK wide. Every example on this page runs in our own business.
The work that is not on anybody's job description
Every small business carries a layer of invisible admin that nobody was hired to do and nobody has ever costed. It gets done in the evenings, at weekends, and in the gaps between real work.
- Reading an invoice off a PDF and typing the same numbers into the accounts.
- Copying an enquiry from an email into a spreadsheet, then into a quote.
- Checking whether an order came through, in three different places.
- Chasing the same information from the same supplier every month.
- Building the same report by hand because nothing joins the systems up.
- Remembering. Constantly remembering, because nothing else does.
None of it is difficult. All of it is repetitive, and repetitive work done by a tired person at eight in the evening is where errors come from.
What we actually mean by automation
The word has been stretched to cover everything from an email rule to a robot arm. Here is the specific thing we build, and the specific thing we will not.
An automation watches for something happening, decides what to do about it based on rules you set, and then does it across the systems you already use. That is the whole shape. The trigger might be an email arriving, a form being submitted, a payment clearing or a date passing.
The important part is the deciding. A simple rule handles the predictable cases. Anything unusual gets flagged and routed to a person with the context attached, rather than being pushed through and quietly getting it wrong. Where that line sits is the single most important decision in any build, and it is your decision rather than ours.
What we will not build is something that runs unattended on work where being wrong is expensive. Paying an invoice, committing to a price, sending anything contractual to a customer. Those get prepared automatically and approved by a human, because the cost of an error there dwarfs the time saved.
Where AI fits, and where it does not
Plenty of useful automation involves no AI whatsoever. A rule that copies a form submission into a spreadsheet is plumbing, and plumbing is often enough. It is cheaper, more predictable, and easier to fix at midnight.
AI earns its place when the input is messy: understanding what somebody wants from a badly worded email, pulling figures off six suppliers' differently laid out invoices, summarising a long thread. That is the part that used to need a person to read it.
The unglamorous test we apply
Before building anything we ask how many times a week it happens and how long it takes. Multiply those, times fifty, and compare it with the build cost. If the sum does not work, we say so.
A surprising number of automation ideas fail that test. Something that happens twice a year should be done by hand, however satisfying it would be to automate it.
Five things we automated in our own business
Not case studies from a brochure. These run at Manchester PC and have done for a long time.
Supplier invoices into the accounts
An invoice arrives by email, the figures are read off the PDF, a bill is created in our accounting system with the original attached, and anything that does not match is put in front of a person. Six suppliers, six ugly layouts, no retyping.
Marketplace sales reconciled
A sale is detected, the invoice is raised, the transaction is recorded, and the stock position updates. What used to be an end-of-day job in three browser tabs is now something we check rather than do.
Enquiries recognised and routed
Incoming mail is classified, matched to a customer where one exists, turned into a lead with an owner, and acknowledged to the customer. Our inbox stopped being an archaeology site.
Read the case studyQuote approvals moving the job on
A customer approves work in writing, the job updates itself, the next stage begins and the technician is told. No verbal agreements to argue about later.
Quote and job managementRepair triage and quoting
An enquiry is read, classified against our own triage rules, and a quote is drafted from a library of priced work. A human approves every one before it is sent. Days became minutes.
Read the case studyManagement reporting
Customer, job, time and accounting information pulled together into the handful of numbers we actually look at on a Monday, instead of somebody building a spreadsheet.
Why we can say this with a straight face
We built all of this for ourselves before we sold any of it
Manchester PC is a repair business. That is not a heritage story we tell for charm, it is the reason this page exists. Every automation described above was built because the admin around the actual work was eating our evenings, and we had the unusual advantage of being able to build software rather than just buy more of it.
The invoice parser is the one we are most attached to, because it was the most annoying to get right. One supplier's PDF layout was genuinely hostile, the sort of thing where the numbers are not really in a table at all. We could have accepted mostly right and checked everything by hand, which would have defeated the point. It took real engineering to make it reliable, and it now runs without anybody thinking about it.
We also pulled things back. A couple of workflows we automated, watched carefully for a few weeks, and then returned to a person, because the output was not good enough for a business with a reputation to protect. Knowing which is which is most of the skill, and you only learn it by running the things in anger.
So when we quote you for a build, we are not estimating from a framework. We have made these mistakes with our own money first, which is why the discovery step exists and why we will tell you when an idea is not worth the money.
A few things we automated, watched, and deliberately undid. That is not a failure of the technology, it is what paying attention looks like.
Read the whole storyHow a build actually happens
Five stages, and you can stop after the first one with something useful in your hand.
- 01
Discovery
We map what happens now, step by step, including the bits that only exist in somebody's head. You get the map and a fixed quote whether or not you proceed.
- 02
Agree the boundary
What the system may do alone, what needs approval, and what should always go to a person. This is your call and we write it down.
- 03
Build and test
Built against your real data and real awkward cases, not a tidy example. The exceptions are where these things live or die.
- 04
Run it alongside
For the first period it runs while a person still checks, so you can see what it does before you trust it with anything.
- 05
Hand over and watch
Logging you can actually read, alerts when something looks wrong, and somebody to ring when it does.
How to find your own candidates
You do not need us to work out where the time is going. This is the exercise we run in a discovery session, and you can do most of it yourself over a coffee.
Take last week and list every task somebody did more than twice. Not projects, tasks. Reading an invoice. Copying a phone number into a spreadsheet. Checking whether an order shipped. Sending the same reminder to the same sort of person.
Against each one, write roughly how long it takes and how often it happens. Multiply. Most people are surprised, not by the big items but by the small ones: the four-minute job that happens eleven times a week is a whole working day a month.
Now mark each task with whether it needs judgement. Be strict about this. Does it need somebody to weigh something up, or does it just need somebody to be there? The second kind is your automation list, and it is usually longer than people expect.
The three shapes worth looking for
The first is information moving between two systems that do not talk. Anything you type into one place having read it from another is a candidate, and it is usually the easiest thing to fix.
The second is anything that depends on somebody remembering. Follow-ups, renewals, checks, reminders. Humans are unreliable at this and feel bad about it, which is a waste of two things at once.
The third is reading. Extracting a handful of facts out of a document or a message so that something else can happen. That used to require a person and largely does not any more.
Start with the one that annoys you most
Not the biggest, not the most impressive. The one that makes somebody sigh. Irritation is a surprisingly good proxy for repetition, and it also means the first project has an obvious champion who will actually tell you whether it worked.
Getting one thing genuinely finished beats half-building four, which is how most automation programmes quietly die.
What it costs
All prices plus VAT. Automation prices vary wildly across the market because the work does. These are honest starting points rather than a ceiling.
Automation Discovery
£249
Finding out whether any of this is worth doing
- One workflow reviewed properly
- A map of how the process actually runs today
- Recommended automation, in priority order
- A fixed implementation quote
- Credited against any build over £1,000
Simple Automation
From £500 setupFrom £49/mo
One source, one destination, straightforward rules
- A single connected workflow
- Human review where it matters
- Error handling and logging
- Documentation you can read
- Support from £49 a month
Connected Workflow
From £1,500 setupFrom £99/mo
Email, documents and your accounting or CRM system together
- Email and PDF extraction
- CRM or accounting integration
- Multiple actions per trigger
- Exception handling and alerting
- Full logging
- Support from £99 a month
Operational Automation
From £3,000 setupFrom £199/mo
Several systems, custom API work and real complexity
- Multiple systems connected
- Custom API development
- Team permissions
- Complex decision rules
- Dashboards and reporting
- Ongoing optimisation
We publish starting points rather than a full price list because two builds that sound identical can differ by a factor of five once you look at the systems involved. The discovery exists so that the number you get is real.
What we will tell you not to automate
A short list, because a page selling automation ought to be honest about its limits, and because these are the conversations that save people money.
A process that is a mess
If a process is chaotic, automating it makes the chaos faster and harder to see. Tidy it first, on paper, with the people who do it. Sometimes that alone fixes the problem and there is nothing left to build.
We have ended discovery sessions by recommending a change to who does what, at no further cost. That is a better outcome than selling a build.
Something that happens rarely
Twice a year is not a candidate however irritating it is. The build cost will never come back, and every automation carries a small ongoing burden of keeping it working when the systems around it change.
Anything where being wrong is expensive and silent
Paying money, committing to prices, sending contractual documents. These get prepared automatically and approved by a person. The time saved by removing the approval is small, and the cost of one bad automatic decision is not.
Judgement dressed up as a rule
If two experienced people in your business would handle the same case differently, that is judgement, and it does not belong in a rule engine. Automate the gathering of the information and leave the decision where it is.
Data protection applies to all of this. Anything touching customer records means deciding what the system may see, where it is processed and how long anything is kept. We settle that at the start rather than retrofitting it.
Frequently asked questions
Do we need to change the software we use?
Usually not. Most of what we build connects the tools you already pay for. Occasionally something genuinely will not integrate and then you have a real choice to make, but we would rather work with your existing setup than sell you a migration.
How long does a build take?
A simple one is one to two weeks. A connected workflow is three to six. The variable is almost never the code. It is how quickly we can get access to systems and how many awkward exceptions turn up once we look at real data.
What happens when it breaks?
It will, eventually, usually because something it depends on changed. That is what the support fee covers. Everything is logged and alerted, so we generally know before you do, which is the arrangement you want.
Can you automate things in FreeAgent or Xero?
Yes. Accounting workflows are among the most common things we build, because invoice and bill handling is repetitive, high volume and genuinely tedious. What is possible depends on what your package exposes, which we check during discovery.
Is the discovery fee refundable?
It is credited in full against any build over £1,000, so if you proceed it costs you nothing. If you do not proceed, you keep the process map and the quote, which several clients have used to make changes themselves.
Will this put somebody out of a job?
In a small business, almost never. What it removes is the part of the job nobody wanted: the retyping, the checking, the chasing. In practice it tends to postpone a hire rather than replace a person, and it is worth being clear with your team about that from the start.
Who owns what you build?
You do. The workflows, the configuration and the documentation are yours, and if you ever want to take them elsewhere we will hand over everything including the parts that are awkward to explain. We are not interested in building something you cannot leave. A supplier who makes their work deliberately difficult to hand over is telling you what they think of their own service.
What if our systems change next year?
Then the automation needs adjusting, which is what the monthly support covers. Software providers change their interfaces without asking anybody, and something that has run perfectly for eight months can stop overnight because a third party altered a field name. That is normal, it is why the support fee exists, and it is why we log everything rather than assuming things will keep working.
Can you work with our existing IT provider?
Yes, and we often do. Automation usually needs access to systems somebody else looks after, and the sensible arrangement is that they keep doing what they do while we build on top. We would rather cooperate than turn it into a territorial argument that leaves the client in the middle.
What good looks like six months in
Automation projects are easy to sell and easy to abandon. These are the things that separate the ones still running from the ones quietly switched off.
The first sign of a healthy automation is that nobody talks about it. It has stopped being a project and become part of how the week works, in the same way nobody discusses the dishwasher. If people are still checking its output line by line after two months, something is wrong with either the build or the trust, and both are fixable.
The second is that somebody owns it. Not a department, a person who would notice if it stopped. That is usually fifteen minutes a month rather than a role, but if the honest answer is nobody, it will degrade and you will find out from a customer rather than from a screen.
The third is that it has been changed at least once. Businesses move, and an automation that has never been adjusted is either perfect or ignored. It is almost never perfect.
We build in logging you can actually read for exactly this reason. The first time something looks wrong, you want to be able to see what it did rather than take anybody's word for it, including ours.
Where this connects
Automation sits around everything else, taking the admin out of it.
- AI Opportunity AssessmentIf you are not sure where to automate first, start here rather than guessing.
- AI ConsultingAdvice, rules and planning before anything gets built. Advisory session £149.
- Lead capture and follow-upAutomating what happens to an enquiry the moment it arrives.
- Quote and job managementAutomating the actions that move a job through your workshop or your week.
- How the platform fits togetherWhere the automated pieces actually live and how they share data.
- How we sorted our own inboxThe full case study of the email triage described on this page.
- What AI automation can actually doThe plain-English version, with no pricing attached.
Start with the job you hate most
Book a discovery and bring the task that irritates you every week. We will map it, tell you honestly whether it is worth automating, and give you a fixed price. The map is yours either way.
All prices plus VAT. Discovery credited against builds over £1,000.