The implementation cost

Development is only part of the investment. Data has to be prepared, access arranged, tests run and time set aside by the people who know the process. If the data isn’t structured, it may need cleaning before anything is integrated.

If an integration hasn’t been verified yet, I say so in the proposal. It is better to resolve that uncertainty at the start than to discover halfway through the project that something can’t be done as planned.

Explore the idea

Estimate the time genuinely released

Enter your own assumptions. The result deducts review time and exceptions.

  1. 01Current volume
  2. 02Review needed
  3. 03Exceptions

Try your own assumptions

Monthly hours released = tasks × (current minutes − review minutes − exception rate × extra minutes) ÷ 60.

Illustration: 200 tasks × (10 − 2 − 10% × 5) ÷ 60 = 25 hours/month. Not a forecast: excludes setup, maintenance, tool costs and other delays. Validate with a real sample.

The running cost

Then there are recurring costs: infrastructure, licences, AI model usage and support. Many grow with volume, so it is worth estimating how many operations the system will process.

One cost is rarely included: the time your team spends reviewing results and handling exceptions. Add to that the adjustments needed whenever one of the integrated systems changes.

Time released is not money saved

The usual calculation multiplies time per task by the number of tasks. That shows the current effort. To know the time genuinely released, subtract the time that will still go into reviewing and correcting.

That time doesn’t automatically become savings or sales. It may translate into better service or a more sustainable workload for the team. Both goals are valid, but it is worth stating which one you are after, because the calculation changes substantially.

Two scenarios with visible assumptions

I compare a moderate-use scenario with a higher-volume one, with every assumption in view. If a single value reverses the conclusion, that value must be verified before deciding.

Once the system is running, I review the real data. Extending the automation to other processes makes sense when the observed usage and cost are consistent with the company’s goals, not because the initial estimate said so.

If you want to run this calculation with one of your processes, here is how I work, or book a meeting and we’ll look at your case.