A practical framework for working out whether a workflow automation is worth building — and how to present the business case internally.
Every automation project starts with the same question: is this worth building? The answer requires actual numbers, not assumptions. Here's the framework we use — with the real figures from recent client projects.
Start by measuring the current state accurately. The three inputs you need:
Hours per week spent on the process (broken down by who does it). Minutes per transaction × transactions per week usually gets you here. Be honest about total time including rework, exceptions handling, and checking for errors.
Fully-loaded hourly cost of the people doing the work. Salary is not fully-loaded cost. Include employer NI contributions, benefits, office space allocation, and management overhead. For UK businesses, a £35,000 salary typically has a fully-loaded cost of £48,000–£52,000.
Error rate and cost of errors. What percentage of transactions have errors? What does fixing an error cost in time and, where applicable, in financial terms (refunds, penalties, reputational damage)?
Example from a real project: A finance team was spending 22 hours per week on invoice reconciliation. Fully-loaded hourly cost: £26. Weekly cost: £572. Annual cost: £29,744. Error rate was 3.2%, with average correction time of 45 minutes. Annual error correction cost: £4,160. Total annual cost of current process: £33,904.
This is what we charge for the automation, including discovery, build, testing, and any third-party API or infrastructure costs. Get a fixed-price quote, not a time-and-materials estimate. T&M automation projects reliably overrun.
For the invoice reconciliation project: £18,000 build cost, £3,600/year ongoing maintenance and hosting.
After automation, some human time remains — reviewing exceptions, handling edge cases, monitoring the system. Be conservative about how much time you'll save. We typically model 80% time reduction as a realistic outcome, not 100%.
Post-automation finance team time on this process: 2.5 hours per week (exception handling and oversight). Annual cost: £3,380. Error rate post-automation: 0.03% (caught by ERP validation before posting). Annual error cost: effectively £0.
Annual saving = (Pre-automation cost) − (Post-automation cost + Ongoing maintenance) Annual saving = £33,904 − (£3,380 + £3,600) = £26,924
Payback period = Build cost ÷ Annual saving = £18,000 ÷ £26,924 = 8 months
Year 1 net benefit: £26,924 − £18,000 = £8,924 Year 2 net benefit: £26,924 Year 3 net benefit: £26,924 3-year total net benefit: £62,772
Automation is worth building when: the process is high-volume and repetitive, the rules are consistent enough to encode, and the time cost of the current process exceeds the build cost within 18 months.
It's usually not worth building when: volume is low (fewer than 50 transactions per week), the process changes frequently, or the exceptions are so complex that human judgment is genuinely required most of the time.
The 18-month payback threshold is our working rule of thumb. If the numbers don't support a payback within 18 months, the business case is weak. Below 12 months is compelling. Below 6 months is a priority.
Two things that help: show the three-year number, not just year one. Automation has a step-cost structure — the build cost is upfront, but the savings compound. And show the non-financial benefits separately: reduced errors, faster cycle times, staff redeployment to higher-value work. These matter to leadership even when they're hard to quantify precisely.