What “loop engineering” looks like inside a small Suffolk business — and the moment a quick AI fix needs to grow up.
Every Monday morning used to start the same way for me: half an hour trawling the government’s Contracts Finder site for tenders worth bidding on. Necessary work. Boring work. The kind of work that quietly eats a diary.
It doesn’t happen any more. Not because I stopped doing it — because I don’t do it. An automated process runs every week, searches the tender database, filters out the noise, removes duplicates, and drops a short summary in front of me. My Monday now starts with a two-minute read instead of a thirty-minute trawl.
There’s a name for this that’s been doing the rounds in the software world lately: loop engineering.
From typing prompts to designing systems
If you’ve used ChatGPT or Claude, you’ve done what the industry calls prompting: you type a request, the AI responds, you type another. You’re in the loop for every step.
Loop engineering is the next stage. Instead of prompting the AI by hand, you design a repeating cycle it runs on its own: find the work, do the work, check the result, record what happened, go again. You stop being the operator and become the designer.
That shift matters for SMEs more than most of the commentary admits, because the tasks that drain small businesses aren’t one-off questions — they’re the same jobs, done the same way, every day or every week. Chasing overdue invoices. Sorting the inbox. Checking a source for new opportunities. Exactly the shape of work a well-designed loop handles beautifully.
What this looks like in practice
Two real examples from inside our own business.
The weekly tender scan. Every week, an automated process searches Contracts Finder for public sector software opportunities. The government’s keyword matching is broad — most raw results are irrelevant — so the process filters them properly, removes duplicates, and produces a short, readable summary. Revenue-hunting that happens whether I remember to do it or not.
The daily inbox sort. Each morning, another process files the previous day’s emails into client folders and gives me a short report of what came in and what needs attention. Small on any given day. Over a year, it’s days of reclaimed time — and nothing important sits unread in a pile.
Neither of these needed a big budget or an enterprise platform. That’s the genuinely exciting part of this shift: this kind of automation is now within reach of a ten-person business, not just a corporation with an IT department.
But here’s the part of the story most of the excitement skips over.
The first version was vibe coded — and that was fine
If loop engineering is the trend of the moment, “vibe coding” was the one before it: describing what you want to an AI and letting it write the code, without worrying too much about how it works underneath.
I’ll be honest — the first versions of both of my loops were exactly that. Described what I wanted, got working code, ran it. Quick, cheap, and it worked.
I want to be clear: that’s not a criticism. Vibe coding is brilliant for what it’s for. It’s the fastest way ever invented to prove an idea. If you’ve knocked together a quick AI tool for your business, you’ve done the modern equivalent of building your first spreadsheet — a sensible, low-cost first step.
But I’ve spent years helping businesses that built themselves on spreadsheets, and I recognised the pattern immediately. Because my vibe-coded loop did what vibe-coded solutions always eventually do.
It failed. Silently.
The early version of my inbox process worked through a web browser, the way a person would. One day, a security prompt appeared that it didn’t expect. It didn’t crash. It didn’t send an alert. It just quietly stopped — and kept quietly stopping for four days before I noticed the reports had gone stale.
Four days of an inbox I believed was being handled, and wasn’t.
The moment a quick fix needs to grow up
Nothing was lost that week. But it crystallised something I say to clients about spreadsheets all the time, now applied to AI: the problem is never that the quick solution exists. The problem is the moment your business starts depending on it without noticing.
The fix for my inbox loop wasn’t a cleverer prompt. It was engineering. The rebuilt version talks directly to the email system through a proper, secure interface rather than pretending to be a person at a keyboard. It knows exactly what it’s allowed to touch. It has defined failure conditions — if something unexpected happens, it stops loudly and tells me, instead of stalling in silence. And it keeps a record of every action it takes.
If that list sounds familiar to anyone who’s worked in a regulated business, it should.
This is where the quality-management people start nodding
Maly is certified to ISO 9001 and ISO 27001 — the international standards for quality management and information security. Strip away the paperwork and ISO 9001 is built on one idea: plan, do, check, act. Define the process, run it, verify the result, improve it. It is, quite literally, a loop — one that’s been keeping businesses honest since long before AI arrived.
And ISO 27001 asks the questions that separate a robust automation from a risky one: What data can this thing access? Who authorised that? What happens when it fails? Can you show me a record of what it did?
A well-engineered AI loop and a well-run quality system are the same discipline wearing different clothes: a clear goal, a way to verify the work, an explicit plan for failure, and an audit trail. That’s not bureaucracy getting in the way of the exciting new thing. That’s the difference between automation you can trust and automation you have to hope about.
It’s also why “just get the AI to do it” is only ever half an answer — especially if you’re in financial services, pharma, or anywhere a regulator might one day ask how a decision got made.
The honest takeaway
So here’s where I’ve landed, and it’s the same place I always land. Vibe-code freely. Experiment. Prove your ideas cheaply — it has never been easier. But know the signs that a quick fix has quietly become critical infrastructure: money moves through it, customers depend on it, compliance touches it, or you’d genuinely struggle for a day without it.
That’s the point where it needs verification, failure handling, security boundaries, and a record of what it did. The same point at which a spreadsheet needs to become a system.
The question worth sitting with: what’s the repetitive weekly job in your business that could be running itself — and what’s the AI experiment you’re already quietly depending on?
If either question made you pause, it’s worth a conversation.
We’re Maly IT Solutions — based in Suffolk, working with SMEs across East Anglia. We help businesses automate the right things, the robust way — and we run our own business on the same loops we build for others.
Free 30-minute call, no commitment — just a practical look at where automation could genuinely help, and where your current tools might need shoring up.
📧 hello@maly.co.uk 📞 01473 934672 🌐 maly.co.uk