Calculate AI value after review and running costs

Count all the time

Record current minutes per case and the assisted work plus human review required afterwards. Multiply the difference by the monthly volume and adoption fraction. Subtract maintenance effort. Do not assume every case will use the pilot.

Separate capacity from cash

Released hours valued at an hourly rate are potential capacity. Salaries do not automatically fall when time is released. Cash savings require a genuinely avoided expenditure; additional revenue requires evidence about what the team does with that time.

Fictional worked example

At 200 monthly cases, 15 current minutes, 4 assisted minutes and 3 review minutes, full adoption releases 26.67 hours before maintenance. After 2 maintenance hours, time value at AED75 per hour is AED1,850. Deducting AED400 monthly costs gives AED1,450 capacity value. AED5,000 setup divided by that figure gives approximately 3.45 months of illustrative capacity-value payback, not a guaranteed cash return.

Test the weaker scenario

At 50% adoption with the same assumptions, net hours fall to 11.33 and net capacity value to AED450. Payback becomes 11.11 months. Test high review effort and higher recurring costs as well. Replace estimates with measured pilot data before an investment decision.

Measure the current process honestly

Observe several representative cases rather than selecting the slowest example as the baseline. Record preparation, lookup, drafting, checking and correction time. Note whether the work is active effort or elapsed waiting. An AI draft may reduce active writing time while leaving approval delays unchanged. Mixing the two can make an estimate difficult to interpret.

Include adoption and exceptions

A pilot may initially cover only some cases. Use an adoption fraction that reflects the proportion likely to use the workflow, and keep exception handling in the estimate. If a case needs a person to start again, that effort belongs in the assisted or review time. Maintenance includes updating sources, checking outputs and handling failures. It should not disappear from the model because the first demonstration was smooth.

Check the full cost boundary

Recurring costs can include model usage, software, support and storage. Setup can include development, testing, team training and migration. Compare figures on a consistent tax basis. If provider or supplier costs are unknown, keep them labelled as estimates and test a range. Do not represent a time-value calculation as a confirmed return on investment.

Translate capacity into a business plan

If the pilot releases hours, ask how those hours will be used. A team might handle more enquiries, reduce an existing backlog or spend more time on complex cases. These are potential uses of capacity; revenue or avoided hiring needs evidence of its own. A cash-saving scenario should identify the expense genuinely avoided, who confirms it and when it becomes real.

Use three scenarios

Create a cautious case with lower adoption and higher review effort, a base case using current evidence and a favourable case whose assumptions you can explain. If the decision only works in the favourable case, that is useful information. Show each input alongside the result so a reviewer can understand what changed. Avoid presenting a single large savings number without its assumptions.

Replace estimates after a controlled pilot

Collect measured time per completed task, correction rate, usage costs and adoption during the trial. Compare the same type of cases before and after, and explain changes in volume or scope. Review whether maintenance was understated. The calculator should evolve with evidence; it is a decision aid, not a sales promise. Keep the dated assumption export with the pilot record so later comparisons are meaningful.

Before you use the result

• Separate approved facts from assumptions.
• Identify missing information and the person who owns it.
• Check prices, promises and consequential fields.
• Review the work in the format the audience will actually see.

Questions to ask next