Audit the Process Before You Automate It

Every week another business decides this is the year it automates the admin or adds AI to the back office. Good instinct. The mistake comes in the very next step, when someone picks a process, wires a tool straight onto it exactly as it runs today, and switches it on. Two questions never got asked: is this even the process worth automating first, and is the process itself any good? If it was quietly bloated or broken, and most are, nothing has been improved. You have simply spent money making the wrong thing, or a poor version of the right thing, run faster. Automation does not ask whether the work is worth doing or done well. It just does the work. Which is why the audit has to come first.

Most processes were never designed

In years of auditing processes across different businesses and departments, one thing showed up every single time. There are always steps that are still done today for no reason anyone can give you, beyond the fact that this is how we have always done it. That is not a criticism of the people doing the work. It is simply how processes come to exist. Nobody sits down and designs the perfect process on day one. Something simple gets set up, the business changes, a problem appears, so a check gets bolted on to catch it. Then another. Years later, the way the work flows is the sum of every patch anyone ever added, and the people running it inherited it rather than built it.

Ask why a particular step exists and the honest answer is often a shrug. That is the process a builder in a hurry will automate: not the right one, just the current one. It is the same trap as pouring money into the tools around a spreadsheet without ever questioning what you built around it.

You pay for every step you automate

Here is what it actually costs you. Picture a process that grew to fifteen steps over the years when the job really only needs five. Automate it exactly as it stands and you are building, paying for and maintaining all fifteen. Every extra step is more to build up front, one more thing that can break, and one more thing for someone to understand and manage later. You have not just automated the wrong thing, you have automated the most expensive, most complicated version of it.

The ten steps you never needed do not disappear either. They sit inside the system, being run over and over. To be clear, that is not the same as hiding them. We build automations you can see and follow, not black boxes, so you keep full visibility. The problem is simpler than that: you are now paying to build and run machinery you did not need in the first place. Right-size the process before you build, and the automation comes out cheaper, simpler and far easier to live with.

Every step is there to manage a risk

Here is the lens an auditor brings that a tool-builder usually does not. Every step in a process is there to manage a risk. Every risk is held in check by one or more controls, and some of those controls are critical, meaning that if they fail something genuinely bad happens. Many of them are not critical at all.

What you find, again and again, is checks that were added years ago to catch a problem that no longer exists, or to double-check a person who left in 2019. They survive out of habit. Once you understand which controls actually matter and which are just there, two things become possible. You can strip out the checks that no longer earn their place, which speeds the work up before a single line of automation is written. And you know exactly which controls have to survive the change, so the important safeguards get carried across instead of being accidentally engineered out. Get that wrong and you can automate away the one check that was keeping you safe.

What an audit actually does

An audit sounds formal. In practice the first part is simply a conversation. Someone independent, who is not tied to any of your current ways of working, sits with the people in your business who actually do the job and maps how the work really flows, not how a policy document says it should. From that you can see what each step is for, what risk it manages, and where the process is fragile, which is usually anywhere a person is rekeying data by hand or making a judgement under time pressure.

Your business does not have one process, it has dozens. You cannot audit them all at once and you should not try. You start where the value and the risk are highest. Sometimes that is simply the process you already know is painful. Often it is somewhere you have not thought to look, which the conversation surfaces: the manual step most likely to go wrong, the one nobody really owns. The point is you do not need to know in advance which of your processes is the one to fix. Working that out is what an AI and automation strategy audit is for.

Optimise first, automate second

The order is the whole point. Talk and map, understand the risks and controls, cut what is not needed, and only then automate the version you are left with. Sometimes that final step is a simple automation. Sometimes it is worth adding AI, where there is genuine judgement or messy information a model can handle well. But that decision comes last, once you actually understand the work, not first because AI is the thing everyone is talking about.

Optimising does not always mean fewer steps. Sometimes a process is already lean and there is nothing worth cutting. And because an automated step is cheap and never forgets, you can often afford to add things a person could not reliably do by hand, like notifying three people the moment something happens instead of hoping one of them remembers to. The goal is the right-sized process, not the smallest one.

This is also the moment the data question matters. If you are going to put business information through an automated system or an AI model, the same audit discipline asks where that data goes and whether it should ever leave your control. A process handling client or financial data might be fine to automate and a poor fit for sending to a third-party AI service. Understanding the process tells you that before you commit, not after.

Get an auditor on your side

Here is the uncomfortable bit for our own industry. Plenty of people can wire two tools together. Far fewer can actually audit a process, because auditing is a distinct skill: knowing what to ask, what to ignore, and how to tell a critical control from a habit. It is the skill I spent years building before I ever automated anything, and it is the part that protects you.

So the honest offer is this. Even if you never have me build the automation, an independent audit first tells you where the real value is and what to fix before you spend a penny on tools. And if you have already chosen someone else to build it, get an auditor on your side of the table anyway, to make sure what you are being sold actually fits your business rather than the seller, the same instinct behind asking hard questions before you sign. The build is the easy part. Knowing what to build, and what to leave well alone, is the work.

Curious whether your process is ready to automate, or whether it needs an audit first? Take the two-minute assessment, or email [email protected] and we will start with a conversation.