The headcount fiction
Almost every automation ROI model starts with headcount. The logic is: this workflow takes 3 people 4 hours each per day, so automating it saves 1.5 FTEs at ~$65K per year, therefore the ROI is $97K annually. Then it gets blessed by finance, included in a board deck, and the automation gets approved.
Then nobody gets fired. Because they were never going to get fired. The three people move on to other work that was previously backlogged, or their teams absorb them into adjacent roles. The company is not worse off, since capacity really was freed, but the $97K does not appear in any P&L line. Finance asks where the savings are. The automation team says "it is in capacity." Finance is not impressed.
The headcount model gets used because the RFP asks for it. It is not a useful number. Here is what is.
Watch out
The three numbers that matter
Build the case around three measurements, and leave headcount out.
1. Cycle time reduction
Measure the end-to-end time for the process today. An invoice that arrives on Monday and gets posted to the ERP by Friday has a 4-day cycle time. After automation, that same invoice is posted within 2 hours. The value is in cash flow visibility, early payment discounts and supplier relationships. For a company processing around 500 invoices a month on net-30 terms, moving from a 4-day to a 2-hour cycle time puts early payment discounts within reach that it currently misses. At an average invoice size of $8K, even a small share of those is real money.
2. Error reduction cost
Manual processes have error rates. Finance teams rarely track them explicitly, but they show up in: rework hours, late payment penalties, duplicate payment writeoffs, and time spent on reconciliation exceptions. Pull three months of bank reconciliation exceptions and categorize them. Data entry errors usually make up a large share. That cost is concrete, and automation removes most of it.
3. Growth capacity without linear headcount growth
This is the real number and it is also the hardest to put a dollar value on pre-automation. Ask the question: if our transaction volume doubles in 18 months, what does that cost us today versus after automation? A company adding a second entity or entering a new market without automation has to hire proportionally. With automation, the marginal cost of additional volume is near zero up to a threshold. That capacity is worth the hires you defer, but only if the growth actually happens.
The only honest version of automation ROI is: here is what it costs us today, here is what changes, and here is the specific business outcome we can trace to that change. The headcount we theoretically freed is not on the list.
How to build the model
Start with a time study. Measure it. Do not estimate it. Have someone log the steps of the target process for two weeks. This sounds tedious and it is, but the numbers that come out of it are defensible. A process the business thinks takes 2 hours a day often takes closer to 3.5 when you count the interruptions, the error corrections, and the ERP navigation time.
Then separate recoverable costs from capacity gains. Recoverable costs are things that go down immediately and show up in P&L: error-related penalties, rework hours billed at fully loaded rate, license costs for manual tools being replaced. Capacity gains are real but only materialize if the business grows into them or if management makes an active decision to redeploy staff.
On the cost side, be honest about implementation. An AP automation for a company processing 600 invoices a month, integrated with the ERP and its approval flows, is a five- or six-figure implementation plus yearly support and model maintenance. Get real quotes and put them in. A model without these numbers is a pitch deck.
The payback period on a well-scoped automation project, counting only hard costs, is more often 18–24 months than 6. If someone is promising you 6-month payback using only hard costs, ask them to show you the error rate data and the cycle time measurement they used. There will not be any.
The question finance will ask that you need to answer
Finance will not approve automation based on capacity. They will approve it based on numbers that appear in a report. So the ROI model needs a line that finance can validate in practice.
The approach that has worked for us: identify one specific, measurable cost that the automation eliminates. Late payment penalties on invoices that miss approval windows are common and usually easy to pull from the ERP. Rework hours on reconciliation exceptions can be estimated from support tickets or exception reports. Pick the number you can defend with three months of data, build the ROI case around that number alone, and treat everything else as upside.
Our take
After the automation is live, measure actual outcomes against the model. Every 90 days for the first year. This sounds obvious, and it rarely gets done. The ones who do are the ones who get budget for the second automation project.
Related reading