There is a laptop on a shelf above the parts washer at Marlow Glazing, plugged in, lid shut, four years old. At 4:40 every morning it wakes up, reads the overnight emails, matches delivery notes against purchase orders, flags the ones that disagree by more than $5, and puts a list on the office printer. The staff call it Rita, after the scheduler who built it, who is not a programmer and who did not tell anyone she was doing it until it had been running for two months.
Marlow has 41 employees and fits windows. Nobody there has ever used the word automation in a sentence about Rita. There was no announcement, no deck, no consultant, and no redundancies — the real Rita still works there, and has, by her own estimate, got about six hours of her week back, which she spends chasing suppliers, a task that turns out to require a person with a grudge.
The unglamorous middle
Almost none of this touches the thing the company actually sells. Nobody at Marlow is automating a window. What is going is the paperwork sediment that accumulates between a customer saying yes and money arriving: retyping the same address into three systems, checking that the invoice matches the delivery, chasing a signature, working out which of yesterday’s eleven voicemails were the same person calling twice.
The Sherwood Index, which asked 1,900 firms under 100 staff what they had automated in the past two years, found the median answer was three tasks, all administrative, none customer-facing, and that 71% of the respondents had not classified any of it as a project. It simply happened, usually on a Thursday, usually because someone got sick of doing it. Sherwood’s more useful finding is who built them: 44% were built by a person whose job title had nothing to do with technology.
It isn’t a system. It’s a favour I did for myself that other people started relying on. If you want to call it a system you can pay me differently.
Why nobody says anything
The silence is not modesty. It is three separate calculations, and they all point the same way. Say the word out loud in a building of 41 people and everyone in it starts wondering which of them is next, regardless of what you meant. Say it to customers and they hear cheaper, which is not what Marlow charges. And say it to yourself and you have created a project, with a scope, an owner and a review — which is the surest known method of killing a thing that currently takes four hours to build and works.
Asked to list the quiet machines, the founders interviewed produced remarkably similar inventories:
- Something that reconciles two lists that should agree and don’t — invoices to deliveries, stock to counts, hours to jobs.
- Something that turns an email into a row in a spreadsheet, because the customer will not stop emailing and the spreadsheet is where work lives.
- Something that sends the third reminder, which is the one everybody feels awkward about and nobody sends on time.
- Something that renames and files photographs from the site, a task universally described as the worst hour of the week before it went away.
- One thing that is frankly a bit much — at Marlow it drafts the quote covering letter — and which every founder mentions last, in a lower voice.
The failure mode is not technical, and it is entirely predictable. These things have a bus factor of one. Sherwood found 38% of the firms surveyed had at least one automation that only one employee understood, and 12% had one nobody understood, still running, still trusted. A cold store in the next county was reported to be relying on a script written by a warehouse manager who left in 2023; it works, it books the couriers, and there is a written rule that nobody touches the server. That is not a story about technology. It is a story about a company with a load-bearing ghost.
The honest accounting is smaller than the pitch anyone would write for it. Marlow’s owner reckons Rita and its two siblings save the business somewhere near 30 hours a week. That is most of a person. He did not remove a person, because the people were already behind, and what he bought was not payroll — it was the ability to take an eighth crew on this year without hiring anyone to do the filing for them. It shows up in the accounts as revenue, not cost, which is why the finance director spent a year unable to find it.
None of the founders here think they are early. They think they are doing what a competent office manager in 1994 did with a macro, in a period when everyone else has agreed to have opinions about it instead. Rita Fenn has been offered a title twice and declined both, on the grounds that a title comes with meetings, and meetings are the one thing she has never found a way to automate.
