what does a manual step cost a year?

the conversation about ai in commerce and manufacturing is almost entirely about sales. more conversions, better recommendations, a bigger basket.

it's a real conversation. it just isn't where the first money is.

most people can say what a system costs. fewer can say what it costs to do without one. not because nobody has thought about it, but because the cost is never written down anywhere. it's spread across twelve months, four departments and a number of people who each spend twenty minutes on something they assume is part of the job.

the four steps we find most often

1. kpi follow-up. figures are pulled from the budget in one place and from the accounts in another. they're put together in a spreadsheet, and the comparison is done by hand. every month, by the same person.
2. the handover between marketing and sales. a lead changes hands, and the details are typed in again somewhere else. sometimes only half of them.
3. following up on production and deliveries. the status lives with whoever called last. if the customer needs an answer, someone has to ask someone.
4. invoicing. the work is done, but it has to be found, described and approved before it can be sent.

the four have something in common, and it's the most important point here: they all sit in the seam between two functions. budget and accounts. marketing and sales. production and customer. delivery and finance.

that's why they stay

a step that belongs to two departments is owned by nobody. there's no one whose budget improves by removing it.

and nobody calls to offer to remove your manual order entry. plenty of people call to offer growth.

the sum

the second half is harder to get excited about and considerably easier to calculate, because the numbers are already in your own operation.

times per year × minutes per time × hourly cost.

three numbers someone in the building already knows. the first is in the system, the second the person doing the work can answer straight away, and the third is in the salary budget.

an example

a company sends 4,000 invoices a year. before each invoice can go out, someone has to find the work, describe it and get it approved.

400 hours is ten working weeks. that isn't a saving you discuss at a management meeting. it's a position.

three things the number doesn't show

the errors. a manual step has an error rate. an invoice that is wrong costs more to correct than to avoid, and it costs something with the customer too.

the delay. the invoice sent three weeks later than it could have been is three weeks of cash flow. that number isn't in the sum above, but it is in your overdraft.

the dependence on one person. the step that sits with a single person isn't just a cost. it's a risk, and it becomes visible the day that person is off sick or hands in their notice.

the most expensive mistake is to automate it

a manual step is almost always a relic. a system was replaced five years ago, and someone started typing in one place and copying to another.

put a robot on that step and you've made unnecessary work faster. you've also made it permanent, because now something has been built on top of it.

the order is map, remove, simplify, and automate last. that order isn't technical, it's commercial, and it gets skipped every time the tool is chosen before the business is understood.

when it doesn't pay

when the process doesn't look the same every time. and when the work shouldn't be done at all.

the last one is worth checking first. the cheapest automation is the process that gets abolished, and a surprising number of reports, approvals and registrations exist purely because they always have.