Automation
The £-per-hour case for automation
You don't need an AI strategy to start automating. You need a simple sum — and the discipline to point it at the right work first.
Most automation projects stall in the same place: a vague sense that "we could automate stuff" that never turns into a business case. The fix isn't a strategy deck. It's arithmetic.
The sum
Pick one repetitive process. Estimate four numbers:
- How many people do it
- How many hours each spends on it per week
- Their fully-loaded cost per hour
- The share of that work that's genuinely rules-based (automatable)
Multiply them out across a working year and you have the prize: hours returned and pounds returned, per year, for that one process. It's a back-of-envelope figure — but it's usually enough to see whether the juice is worth the squeeze. Our automation ROI calculator does exactly this sum in ten seconds.
Why per-hour framing beats "AI strategy"
"Should we adopt AI?" is unanswerable — too big, too abstract. "Is it worth automating the 6 hours a week each of our 8 analysts spends assembling the same report?" is a decision a manager can make before lunch. Framing automation as £-per-hour turns a philosophical debate into a prioritised list.
Automation isn't cheaper than a person by default. It's cheaper when the work is repetitive enough that the build pays back in returned hours.
Point it at the right work first
Not all hours are equal. The best first candidates are high-volume, rules-based, and painful — reconciliation, report assembly, data entry, chasing approvals. Avoid work that's genuinely judgement-heavy or rare; the payback isn't there and the risk is higher. Automate the boring, repetitive slice where reliability is easy to prove, bank the hours, and reinvest them in the work only people can do.
Measure hours returned from day one
The trap is treating automation as a one-off saving you never verify. Track hours returned per process from the start. It keeps the programme honest, funds the next automation from the last one's savings, and turns "we bought some AI" into "we returned 10,000 hours to the business this year." That's a number a board understands — and it started with a four-number sum.