Business process automation usually starts by writing the process down

An owner of a healthcare group with several clinics told me what he wanted, and it was not growth and it was not cost.

One person ran finance and administration for the whole group. She was very good. And almost all of how it worked lived in her head rather than anywhere written down.

What he wanted, in his words, was a digital version of her.

What he was actually asking for

He was not asking for AI. He was asking for the business to stop being one person's holiday away from a problem.

That comes up constantly. Somebody starts a conversation about business process automation and what they are really describing is a process nobody has ever documented, which has been quietly waiting for a reason to get fixed.

AI is frequently just the excuse that finally gets the writing down done.

Which is worth knowing before you go looking at platforms, because the writing down is the part that creates the value, and it is the part you can do without spending anything. The same goes for software, so check what your existing systems can already do before you buy anything new.

Why the software is rarely the constraint

The Department for Science, Innovation and Technology found something that ought to be more famous than it is. Inside businesses that have adopted AI, only around 30 per cent of staff actually use it. More than three quarters of adopting firms have under half their workforce using the tools at all.

So the licence got bought and most people carried on as they were.

The ONS, surveying more than 38,000 businesses, found the same pattern. The average firm using AI runs 1.6 AI technologies, and only one in ten describes its use as extensive.

That is not a software problem and no platform fixes it. A business process automation company can build you something excellent and it will sit unused if the people whose job it changes were not part of deciding what it should do.

I would rather sell a business two days of training and a small automation than a platform, because I have watched which one is still running in six months.

What to write down

You do not need a flowchart or a system. You need four columns on one page.

What happens. Who does it. What they need before they can start. What breaks if they are off sick.

Do that for your three most important recurring processes. The quote, the invoice run, the monthly report. Whatever yours are.

That last column is where the interesting stuff is. Anything with one name in it is a risk you are carrying whether or not you ever automate anything.

What you could do on Monday

Pick the person in your business who knows the most that nobody else knows.

Sit with them for an hour and get them to talk you through one process, while you write it down. Not them writing it. You writing it, so the questions get asked.

An hour. One process.

You will find two or three steps they do automatically that nobody else knows exist, and at least one that made sense in 2021 and does not now.

That hour is the actual first step of business process automation. Everything after it is easier and cheaper because of it.

The honest limit

Writing the process down does not mean the process is good.

Sometimes what you find is that the thing works well and the only issue is that it is undocumented. Then you write it down, you sleep better, and you do not automate anything. That is a legitimate outcome and I have had it happen more than once.

The other honest bit. A digital version of the person who knows everything is not a real thing. You can capture what she does, and you can automate the mechanical parts, and you will not capture the judgement she applies when something looks wrong.

That judgement is the reason she is good at the job. What automation buys you is her spending more of her week using it, and less of her week typing.