The full-time equivalent is the number that lands
An annual pound figure is abstract. "One and a half people, doing nothing but this, all year" is not. That's usually the line that makes an owner go and check the hours properly.
What does your manual work actually cost?
Nobody adds this up, because it never shows up as a line on the P&L. It's buried inside salaries you were paying anyway. Put the hours in below and see the annual figure, where it's concentrated, and how much of it AI could realistically take off your hands.
Three things worth noticing in your result
The biggest line is rarely the one you'd guess
Most owners expect data entry to top the list. It's usually chasing, or answering the same questions, because those are spread thin across lots of people and never feel like a task in their own right.
Your first guess is almost certainly low
Ask the person who does the work, not the person who manages it. In every audit we've run, the hours came back higher after we watched the work happen than they did from the first conversation about it.
Questions about the numbers
What counts as manual work?
Anything your team does more or less the same way every time. Retyping information from one system into another, chasing a form or an invoice, writing a reply you've written before, pulling numbers into a spreadsheet, walking a new customer through the same setup steps. If someone could write down the steps and hand them to a new starter, it counts. Work that needs a judgement call each time does not.
What hourly cost should I use?
A loaded hourly cost, not the raw wage. Take the salary, add employer National Insurance, pension contributions, software licences, and the share of overheads that person carries, then divide by their working hours. It usually lands 20 to 30% above the headline hourly wage. If you're guessing, guess a little high: most owners undercount this.
Are the 40, 60, and 80% bands realistic?
They're honest bands rather than a promise. The conservative band is what you get automating only the clearly rules-based work. The middle band is where most builds settle once they've been running a few months, with a person kept on anything needing judgement. The top band needs you to change the process rather than automate it as it stands, which is a bigger organisational ask. Anyone quoting you one confident percentage for what AI saves is guessing.
Why does the number look so high?
Because it's the whole team for a whole year, and nobody ever adds it up. A few hours a week across five people at a normal loaded cost runs into tens of thousands a year without anyone noticing, because it never appears as a line on the P&L. It's spread across salaries you were paying anyway. That's exactly why it survives so long.
Does this mean I should make redundancies?
Not in our experience, and it's not what we sell. Almost every business we work with puts the recovered hours back into work that was already being neglected: following up leads properly, onboarding customers faster, actually answering the phone. The maths here is about capacity, not headcount. If your plan is to cut staff, we're probably the wrong studio for you.
What do I do with this number?
Two things. First, sanity-check it with the people who actually do the work, because your estimate is almost certainly low. Second, work out whether the recoverable figure justifies doing something about it. If it does, our AI audit maps every one of these processes properly, costs them, ranks what to automate first, and gives you a working proof of concept on the top item.
Now find out which bit to fix first
A total is a good reason to act, but it doesn't tell you where to start. Run your processes through the systemisation scorecard, read what repetitive admin really costs for the arithmetic behind this figure, or book a free call and we'll walk through it with you.