Three weeks ago I spoke with an operations director at a mid-sized accountancy practice who had been putting off a conversation about automation because she wasn’t sure what she was buying. “I don’t want to pay someone to tell me I need to pay someone more,” she said. It’s a fair concern, and I hear it often enough that it’s worth addressing directly.
An automation audit is a scoping tool. Its job is to tell you where the time is going, which of those places is worth fixing, and roughly what fixing it would cost and take. Done well, it saves you from commissioning work you don’t need. Done badly, it’s a slide deck that tells you transformation is possible.
Here’s what a genuine one looks like.
What gets reviewed
The starting point is always the same: a list of recurring tasks. Not a strategy document, not a process map. A list, ideally from the people actually doing the work, of things they do more than once a week that feel like they shouldn’t require a human.
From that list, I’m looking for four things:
Volume. How many times does this happen per week or month? A task that happens twice a year isn’t worth touching. A task that happens forty times a week probably is.
Consistency. Does the task follow the same steps every time, or does it require judgment? Automation handles consistency well. It handles judgment poorly. A task that’s eighty percent consistent and twenty percent judgment is usually still worth looking at, because you can automate the eighty percent and flag the rest for human review.
Current tooling. What software is already in use? This matters because the cheapest fix is almost always one that works inside tools the firm already pays for. A Level 1 fix, in my framework, is education: a prompt, a template, a setting nobody knew existed. No new spend required.
Downstream impact. If this task is slow or error-prone, what does that cause further along? A slow invoice approval process isn’t just an accounts payable problem. It affects supplier relationships, cash flow forecasting, and the time the finance manager spends chasing sign-offs.
What the output should look like
A good audit produces a short, ranked list. My version is typically a one-page document with three to five opportunities, each described in plain terms: what the task is, how often it happens, what fixing it would involve, and a rough cost and time estimate.
The question an audit should answer is not “could this be automated?” Almost anything could. The question is “should it be, and at what level?”
Each opportunity sits at one of three levels. Level 1 (education): the fix is already available inside a tool the client has, and I can show them in fifteen minutes. Level 2 (integration): two or more tools need to connect, via something like n8n, Make, or Zapier. Configuration work rather than engineering. Usually a few days and a few hundred to a couple of thousand pounds. Level 3 (custom build): genuine engineering, bespoke logic, four to twelve weeks. Only warranted when the problem actually requires it.
Most audits I run find one Level 1 fix, one or two Level 2 opportunities, and no Level 3 work. That’s typical. Firms that think they need a custom build usually don’t.
How long it takes and what it costs
A full audit for a team of ten to thirty people takes roughly half a day of my time: one hour of interviews, one hour reviewing the tools and any process notes that exist, and two hours writing up the findings. I charge a fixed fee for this. The number isn’t the point here; what matters is that it’s fixed, scoped, and doesn’t commit you to anything further.
If you’re working with a consultant who won’t give you a fixed price for the audit itself, that’s worth noticing.
The thirty-minute version you can do yourself
You don’t need a consultant to identify your top three automation opportunities. Here’s a version you can run in under an hour.
Sit down with whoever manages day-to-day operations and ask two questions. First: what do you do every week that you copy and paste information to complete? Second: what do you chase people for on a regular basis?
Write down every answer. Don’t filter yet.
Then go through the list and mark anything that happens more than ten times a month with a star. For each starred item, note what software is involved.
Now look at your starred items. If any of them involve moving information between two tools you already use (say, from an email into a spreadsheet, or from a form into a CRM), that’s almost certainly a Level 2 fix. If any of them involve doing something inside a single tool that takes longer than it seems like it should, search the tool’s help documentation for “automate” or “workflow.” You may find the answer already exists.
The honest outcome of this exercise, for most firms, is a list of two or three things worth looking at properly. That’s the point. You now know where to focus, and you haven’t spent anything to find out.
What an audit isn’t
It isn’t a commitment to further work. It isn’t a discovery process designed to justify a large project. It isn’t a document full of possibilities. It’s a short, specific answer to the question: where should we start?
If you come away from an audit with a clearer picture of three opportunities and a rough sense of what each would cost, it has done its job. If you come away with a presentation about digital maturity, ask for your money back.
The operations director I mentioned at the start ran the thirty-minute version herself before we spoke. She came to the call with a list of four tasks and a rough sense of which two were worth pursuing. We spent forty minutes confirming her instincts were right and scoping the Level 2 work on one of them. The audit, in that case, was mostly already done.