Automate a Broken Process, Break It Faster

There's a particular kind of disappointment that comes from automating something. You take a job that eats hours every week — the invoicing, the onboarding, the chasing of unpaid invoices — you wire a tool into it, switch it on, and for a fortnight it feels like magic. Then the complaints start. The same mistakes as before, only now there are more of them, they arrive faster, and nobody's quite sure where they're coming from. That isn't bad luck, and it isn't a bad tool. It's what automation does. It doesn't fix a process — it runs it, faster, more often, and with fewer people watching. If the process was sound, that's a gift. If it wasn't, you've just built a machine for making the same mistake at scale. In short, **automating a broken process** breaks it faster. Bill Gates put it plainly enough in *The Road Ahead* that it's worth borrowing: apply technology to a good operation and you magnify what's good about it, but "automation applied to an inefficient operation will magnify the inefficiency." The tool is an amplifier. It has no opinion on what it's amplifying. ## Why it's such an easy trap Nobody automates a broken process on purpose. It happens because the process doesn't look broken from where you're standing — it just looks slow. And "slow" has an obvious fix: make it faster. Automation is sold on exactly that promise. Not "understand your workflow," but "save ten hours a week." One of those is a weekend of unglamorous digging. The other is a button. There's also the awkward fact that fixing a process is real work, and it's usually somebody else's. The flaw in your onboarding might live in a handover between sales and delivery that no one actually owns. Sorting it means a conversation, some disagreement, and a few decisions about who does what. Buying a tool skips all of that. It feels like progress, and it lets everyone avoid the meeting. So the tool goes in, the broken handover stays exactly where it was, and now it's automated. That's the pattern we describe in [Office Heroes & Broken Processes](/insights/office-heroes-broken-processes/). And once it's wired in, it's harder to see and harder to change. A manual mess is at least visible — someone's clearly struggling, and you can watch where it goes wrong. Automate it and the mess disappears behind a clean interface. It looks handled. It's just handled badly, quietly, at volume — and now it's expensive to unpick, because unpicking it means undoing the thing you just paid for. ## What good looks like The businesses that get real value from automation almost always do a boring thing first. They map the process as it actually runs, not the tidy version in their heads. Where does a job start? Every hand it passes through, every wait, every place it doubles back or gets dropped because one person was off that day. Written down, most processes turn out to have two or three steps that exist for no reason anyone can remember, and one or two points where the whole thing quietly falls over. That's the whole point of [process mapping](/services/process-mapping/). That map is where the real work happens, and it usually isn't technical. Often the fix is removing steps, not adding tools — deleting an approval nobody reads, merging two forms into one, giving a single person clear ownership of a handover that used to fall between two. Do that, and something useful happens: a good chunk of the "we need to automate this" pressure just evaporates, because a lot of it was never a speed problem. It was a mess problem. Only then is automation worth a look — on the steps that are left, the ones that are genuinely repetitive, rules-based and high-volume. That's where it earns its keep, and where AI might reasonably come in. On a clean process, an amplifier is exactly what you want — the kind of gain we cover in [AI Workflows: Create Capacity Without Hiring](/insights/ai-workflows-create-capacity/). ## The order that actually works So the sequence matters more than the tooling. Map it. Fix it. Simplify it. Then, and only then, automate what's left — if anything still needs it. That last clause is the one people skip. Sometimes you go through all this and find the automation you were about to buy isn't necessary any more, because the three steps you cut were the whole problem. That isn't a failure. It's the cheapest fix you'll ever make. None of this is anti-automation, and it certainly isn't anti-AI. It's just the right order. A broken process doesn't need a faster engine bolted onto it. It needs fixing. Once it's fixed, an engine might be exactly right — or you might find it runs perfectly well without one. It's the same principle as [Don't Automate Chaos](/insights/dont-automate-chaos/). The verdict on this one: fix the process first. AI is the second move, not the first — and every so often, once you've made the first move, you'll find you don't need the second at all. ## Frequently asked questions ### Why does automating a broken process make things worse? Because automation runs a process faster, more often and with fewer people watching. If the process is flawed, you scale the flaw — more errors, arriving quicker, and harder to trace behind a clean interface than when the work was done by hand. ### Should I map my process before buying an automation tool? Yes. Map how the work actually runs, not the tidy version in your head. Most processes have steps that exist for no reason and one or two points where they quietly fail. Fixing those often removes the pressure to automate at all. ### What kind of steps are actually worth automating? The ones left after you've simplified: genuinely repetitive, rules-based, high-volume tasks on a clean process. On a sound workflow, automation is an amplifier that pays off. On a broken one, it just amplifies the mess. If you're ready to understand where AI can genuinely save you time, reduce friction and create real capacity in your business — without automating a guess — we're happy to talk it through. [Book a discovery call](/contact/) and we'll look at your situation honestly. *Peter Lowe — AI Consultant and Founder, Smart AI Studio.*

This article was written by Peter Lowe. The ideas and opinions are his own; AI was used to assist with drafting and editing.