Will this automation pay for itself?

Record the number of tasks per month, average minutes per task and an agreed hourly cost. Multiply tasks by minutes and divide by 60 to estimate hours. Use representative weeks rather than the busiest day you remember.

Include exceptions. A process that takes five minutes most of the time but an hour when information is missing needs more than a single best-case estimate.

People may still need to review outputs, resolve mismatches and maintain the process. Include that time along with software, AI usage, hosting and support costs.

Released capacity is not automatically a cash saving. If nobody’s paid hours change, the value may be faster service, fewer delays or room to take on more work. State which benefit you are estimating.

Suppose a team handles 400 tasks a month at 10 minutes each. That is about 66.7 hours. If an automated process still needs three minutes of review per task, 20 hours remain and about 46.7 hours are released.

At an assumed £25 an hour, that is approximately £1,167 of monthly capacity. Subtract an assumed £200 a month in running costs and the estimated net capacity value is about £967. A hypothetical £6,000 implementation would take roughly 6.2 months to match that value.

That is a simple estimate, not a cash-return promise. It excludes audit fees, rollout time and any additional training or maintenance. Test a lower-volume month and a higher review-time assumption before deciding.

Continuing the illustrative example, adding a £2,500 audit to the hypothetical £6,000 build gives £8,500 upfront. At £200 a month, running costs add £2,400 over a year: £10,900 in total, before any additional rollout or training costs. Approximately 560 hours of capacity at the assumed £25 an hour would be worth £14,000 over twelve full months of operation.

On those assumptions, net monthly capacity value of roughly £967 would cover the £8,500 upfront cost in about 8.8 operating months. A delayed rollout would push the calendar date back. This is capacity value, not a promise that cash costs will fall.

If volume falls to 200 tasks a month, net capacity value drops to about £383 a month and the same upfront cost takes approximately 22.2 operating months to match. If review instead takes five minutes at the original 400-task volume, the estimate becomes about £633 a month and 13.4 operating months. Use the less comfortable assumptions as well as the optimistic ones.

SME AI audits start from £2,500, team audits from £5,000 and larger organisations from £7,500. Bespoke builds start from £5,000 and are scoped separately.

An embedded AI transformation partnership starts from £2,000 a month with a three-month minimum. The right route depends on whether you need to find the opportunities, implement a defined system or work through a broader plan.

The number of systems is only part of the scope. A workflow with reliable inputs and a supported integration is different from one with scanned documents, inconsistent records and several approval stages. Data preparation, access restrictions, exception handling and a review interface can all add work.

Ask which costs are one-off and which recur. A useful estimate separates discovery, build, testing, team training and handover from hosting, model usage, existing software licences and ongoing support. Check whether your current software plan includes the integration you need before making a commitment.

Give both suppliers the same process, sample types and approximate monthly volume. Ask each to explain what is included, what you must provide, how acceptance will be checked and what happens when an input cannot be processed. A lower build price is difficult to compare if it leaves out review, training or support.

Agree a realistic operating scenario, including busy months and the time a person spends checking outputs. Ask who owns the accounts, code and data, how usage costs will be visible and how you can export your information or hand the system to someone else. Confirm whether quoted prices include VAT and third-party fees.