How to measure automation ROI: saved hours are not always cash savings

Start with a baseline for one workflow

Before changing the work, record what happens today: how many times the task occurs, the hands-on time, waiting, exceptions and rework. Use a representative period and identify which figures are observations and which are estimates. A busy week or a single unusually difficult case should not become the unquestioned baseline.

Keep the unit clear. Minutes per completed order, elapsed time from request to approval and staff hours spent correcting errors answer different questions. The GOV.UK service-benefits guidance explains why a baseline matters; it is public-service guidance, not a forecast of what an SME will save.

Separate recovered capacity from cash savings

If a task takes less staff time, the business has recovered capacity. That time may improve customer service, reduce overtime or allow more work to be completed. It is not automatically money returned to the bank account. A salary continues to be paid unless an actual spending decision changes.

Record a cash saving only when an expense really falls, such as avoidable contractor charges, overtime or an unnecessary licence. Keep improved throughput or a better experience separate until you have evidence of its financial effect. Do not count the same saved hour once as capacity and again as reduced spending.

A fictional worked example

Imagine 600 monthly requests previously requiring 6 minutes of handling each. After a change, normal handling averages 2 minutes, while exceptions require another 8 hours a month. The original handling is 60 hours; the new total is 20 + 8 = 28 hours, leaving 32 hours of recovered capacity. These invented inputs are an arithmetic exercise, not a Veda client result or delivery promise.

At a fictional internal costing rate of £25 an hour, that capacity has an indicative value of £800. It is not an £800 cash saving. If new software and monitoring cost £150 a month, retain that cost separately. If the measured workload or exception rate changes, update the model rather than preserving the first attractive number.

Include the work needed to keep the improvement useful

Account for implementation, testing, staff introduction and continuing operation. A model or connector may require licences, hosting, monitoring and someone to handle exceptions. Staff must also spend time checking output and maintaining the process when business rules change.

For AI-supported work, include the checking needed to make output acceptable. A faster draft that takes longer to verify may not save time overall. The ICO AI guidance is relevant when personal information is involved; resolve appropriate handling before including that workflow in a live trial.

Review observed use and the next decision

After release, compare the same measures with the baseline and explain changes in volume, task mix or staffing. Check whether people use the new workflow and whether unresolved exceptions have moved effort elsewhere. If the evidence is incomplete, record the benefit as unconfirmed.

Use the review to decide whether to improve the workflow further, adjust it or stop. A Business Improvement Partnership gives you ongoing software, AI and consulting time to work through agreed priorities; it does not guarantee a particular return. Start with a free AI Strategy Call, or choose the optional £995 plus VAT Business Efficiency Audit if you want a standalone diagnosis before deciding about delivery.

Sources and further reading

GOV.UK: Measuring the benefits of your service

ICO: Artificial intelligence guidance