Copying data, chasing approvals and correcting records are small tasks. Their cost becomes hard to see when it is spread across people, tools and different moments of the day. Before automating, bring that work into an estimate that can be explained and checked.

The useful question is not how much a tool promises to save. It is how much effort a process consumes today, which part could be avoided and what work would remain after the change. This guide offers an initial operational calculation, not a financial valuation or a promise of returns.

1. Define one process and measure active work

Choose a concrete unit: handling a request, preparing an order or onboarding a supplier. Define its start and end, who participates and what makes a case complete.

  • Volume: cases completed over a representative period; separate campaigns and peaks.
  • Active time: minutes people spend on each step, including checks.
  • Exceptions: cases needing clarification or correction and the additional effort involved.
  • Waiting: elapsed time until resolution, recorded separately from human effort.

A request may wait two days while requiring only twenty minutes of work. Do not turn all waiting time into salary hours. Do not add corrections again if they are already included in average time per case. Measure a sample covering routine and difficult cases, record its limitations and avoid unnecessary personal data.

2. Build a baseline without double counting

Effort hours for the period = volume × average active minutes per case ÷ 60. If rework was measured separately, add only effort not already included. Where roles have different costs, calculate each part separately.

To value those hours, agree an hourly cost with Finance that fits the purpose of the estimate. Document what it includes, excludes and which period it represents. The result describes resources devoted to the process, not necessarily an expense that can be eliminated.

Illustrative example: from minutes to a decision

Invented assumptions to explain the calculation, not customer data: 400 requests a month, 12 minutes of routine work per request, and 40 corrections requiring 10 additional minutes each. Corrections are not included in the 12 minutes.

  • Routine work: 400 × 12 ÷ 60 = 80 hours a month.
  • Additional corrections: 40 × 10 ÷ 60 = 6.7 hours a month, rounded.
  • Estimated baseline: 86.7 hours a month. At an assumed €30 per hour, this represents approximately €2,600 of allocated capacity.

This does not mean automation will save €2,600. If a pilot recovered 25 hours but required 5 hours of oversight, it would release 20 net hours of capacity. At the same rate, that is €600 of valued capacity before licences, support and implementation. Cash savings would arise only if an actual payment were reduced, such as overtime that no longer needed to be paid.

3. Compare with the process that would remain

Automation also creates work. Include checking results, handling exceptions, maintaining integrations, training, support and recovery from failures. Separate initial and recurring costs and use the same time horizon when comparing alternatives.

  • Recovered capacity: hours available for other work; identify that work and who will use the capacity.
  • Cash savings: payments that would actually be avoided, with a reason and an owner to confirm them.
  • Quality and service: errors, lead times or traceability that might improve; measure signals without automatically turning them into euros.

If you value recovered time and reduced overtime, check that the same benefit is not counted twice. Do not attribute revenue to a faster response without evidence connecting the two.

4. Test assumptions before committing investment

Prepare conservative, central and favourable scenarios. Vary volume, avoidable minutes, exceptions, adoption and maintenance. If the decision works only in the favourable scenario, identify the information you need to measure before proceeding.

The GAO Cost Estimating and Assessment Guide recommends stating scope and assumptions, collecting data, analysing sensitivity and risk, and updating estimates with actual costs. It is a methodological reference for project estimates, not certification of this calculation or a business outcome.

Source: GAO — Cost Estimating and Assessment Guide

Our simplified calculation helps frame a pilot. It does not replace a lifecycle estimate or inherently account for financing, tax or discounting. Investment decisions should be checked with Finance and the technical and process owners.

What to bring to a diagnosis

  • A defined workflow, its owner and the period observed.
  • Volume, active times and exceptions, with sources and sample limitations.
  • Current and future costs separated without duplicates.
  • An improvement hypothesis and a pilot with acceptance criteria.

Start with an estimate you can revise. The next step need not be buying software: it could be simplifying an approval, removing duplicate entry or measuring an exception more carefully.

Identify which processes are worth automating

Codiloop can help you define a process and prepare a test with clear data, boundaries and owners.