Skip to main content
Boermans Digital
Main menu

Decision guideAutomation

Build a realistic business case for AI and automation

A recurring task takes five minutes. Multiply that by the monthly volume and automation can look compelling. Yet not every case leaves the work queue. Checks, exceptions and maintenance remain. A useful business case makes that work visible too.

Calculation using fictional figures, not measured savings or forecasts.

Investigate the assumption that could change your decision

The scenarios below show how much human effort might remain. Next, decide which assumption to measure first. Start with the uncertainty that most affects the decision rather than adding decimal places.

On a narrow screen, focus the table and use the arrow keys to scroll.

From observation to decision
UncertaintyEvidence to collect first
Exceptions take longer than expected.Measure their full handling, including investigation and correction.
Released time is spread across employees.Determine which work can be rescheduled and who actually uses that capacity.
Licence or operational costs remain unknown.Request a defined cost estimate and show what it excludes.

Fictional sensitivity test: ten minutes per exception

Keep the middle scenario in this article: 1,200 cases per month, 70% regular cases, one minute of review per regular case and eight hours of fixed operational work. Change only total human handling per exception from six to ten minutes.

The calculation becomes 840 × 1 + 360 × 10 = 4,440 minutes: 74 hours of processing plus eight operational hours, or 82 hours per month. Against the current 100 hours, this leaves 18 hours of potentially released capacity rather than 42. The 24-hour difference comes solely from this fictional assumption; it is not a measured result or financial saving.

Make the post-trial decision explicit

  • Proceed: the agreed quality threshold is met, and costs and remaining work fit the objective.
  • Adjust: a specified uncertainty needs further measurement or a narrower scope.
  • Stop: required quality or manageability cannot be achieved within the agreed boundaries.

Record criteria and open assumptions in the pilot checklist

Our own fictional calculation, checked on 14 October 2026. Volume, time and the share of regular cases are assumptions. The 82 hours already include eight hours of internal operational work; do not count these twice. Add costs and actually usable capacity to the assessment.

Measure the task you actually intend to change

Define a clear start and finish. Count cases over a representative period and measure active handling time. Waiting for a colleague is different from spending five minutes entering data. Both may matter, but they cannot be added together as if they were the same labour input.

Include quieter days, peaks and exceptions. Record whether a case involves normal processing, correction or clarification. Identify work outside the chosen boundary. Otherwise you may measure a broad current process and compare it with only the easiest future step.

A fictional calculation with three scenarios

Assume 1,200 cases per month currently need an average of five minutes of active work each, including existing corrections: 100 hours per month. In this example, future routine cases need one minute of human checking, while exceptions need six minutes of total human handling. Those six minutes include the check; they are not six additional minutes.

The percentages below describe assumed shares of cases using the routine path, not model accuracy. We also assume eight fixed maintenance hours per month. Licensing, implementation and initial setup effort have not been included in this comparison of hours.

On a narrow screen, scroll the table sideways. With a keyboard, focus the table and use the arrow keys.

Fictional monthly example: 1,200 cases, 100 current hours and 8 future maintenance hours.
ScenarioRemaining human effortPotential capacity released
Cautious: 50% routine600 x 1 + 600 x 6 = 4,200 minutes; 70 + 8 = 78 hours100 - 78 = 22 hours per month
Middle: 70% routine840 x 1 + 360 x 6 = 3,000 minutes; 50 + 8 = 58 hours100 - 58 = 42 hours per month
Favourable: 90% routine1,080 x 1 + 120 x 6 = 1,800 minutes; 30 + 8 = 38 hours100 - 38 = 62 hours per month

These figures do not establish a return on investment. Longer exceptions, more maintenance or lower volume change the result. With no routine cases, this example would require 120 processing hours plus 8 maintenance hours: 28 more hours than today. Replace the assumptions with your own observations before drawing conclusions.

Available hours are not automatically cash savings

Scattered minutes do not necessarily become a usable working day. Discuss what other work could be done and whether scheduling or responsibilities would need to change. Serving customers sooner or reducing a backlog can be valuable without immediately reducing payroll.

Multiplying hours by an internal hourly rate places a value on capacity; it does not prove lower expenditure. Claim a cash saving only where an actual cost changes, such as a demonstrated reduction in external support or overtime. Do not count the same hours as both a cost saving and extra capacity.

Want to use your own process hours? The automation calculator shows a time-saving scenario and, optionally, the value of released capacity. Complete the business case with the costs and assumptions discussed in this article.

Use your own process hours in the automation calculator

Include costs across the operating life

  • One-off work: process analysis, configuration, integrations, testing and staff preparation.
  • Recurring charges: licensing, usage, hosting where required and support.
  • Internal effort: checks, exceptions, maintenance, change testing and access management.
  • Transition: temporary parallel running and resolving errors or backlogs.
  • Replacement or exit: moving data, documenting the process and returning to another way of working.

On a narrow screen, scroll the table sideways. With a keyboard, focus the table and use the arrow keys.

Fictional cost example: external expenditure over twelve months. Not market prices or a quotation.
External expenditureCalculationAmount
Setup and integrationOne-off€4,500
Testing and staff instructionOne-off€1,500
Licences12 × €150€1,800
Separately agreed external support12 × €100€1,200
Total in this example€6,000 + 12 × €250€9,000

The amounts are fictional, not Boermans Digital rates. They show only the external expenditure listed. Internal effort, any transition or exit costs and tax effects are not included in this total. External support is a calculation assumption, not a commitment to ongoing support.

The 58 hours of remaining human effort per month in the middle scenario above already include checks, exceptions and eight hours of internal maintenance. Do not add those eight hours again. Count external support separately only when it covers different, separately agreed work.

The potential 42 hours freed per month represent capacity. Avoided expenditure has not been established in this example. Consider a financial saving only where spending actually falls, for example through demonstrably less overtime or external staffing. Do not count the same hours as both additional capacity and a cash saving.

Ask suppliers to explain the assumptions in their quotations and what is excluded. Avoid a precise payback period while material costs remain unknown. Identify the missing information most likely to change the decision first.

Which costs are specific to AI?

AI implementation costs depend on the task, existing software, data and required review. Separate one-off preparation from ongoing use. Check what your current software already offers; a new product is not always necessary.

  • Preparation: selecting and cleaning sources, setting up access, creating test cases and implementing integrations where needed.
  • Use: licences or consumption charges, depending on the product and plan, alongside human review and corrections.
  • Maintenance: keeping sources and instructions current, testing changes and helping staff handle exceptions.

A faster draft can require more checking. Include the total human effort, not just the time the AI takes. The work actually needed becomes part of the agreed scope; this is not a price list.

How should you assess RPA costs?

Robotic Process Automation (RPA) lets software carry out agreed actions in applications. Alongside setup and licences, changes to screens, access rights and process rules may require maintenance. Include failures, manual recovery and ongoing management. Compare this approach with an existing feature or direct system integration before investing.

Discuss the options through AI consulting and record the assumptions to test in the pilot checklist.

Define what the trial needs to show

Agree which observations would support continuing, changing direction or stopping. Use the same process boundaries during the trial as in the baseline. Keep handling time, corrections and exceptions visible separately, and check the quality of the outcome as well.

Give a process owner responsibility for reviewing the figures after introduction. A low-volume month or a period of extra support can be misleading. Update assumptions when evidence changes, but retain the original basis so people can understand why the decision has shifted.

Three useful inputs for an initial conversation

You do not need a finished financial model before discussing the process. An approximate volume, a few examples of manual work and an understanding of the main exceptions provide a useful starting point. The aim is to identify assumptions worth testing, not defend an attractive number.

Which processes are worth investigating?

Which assumptions drive your business case?

We can look at where the time goes, what work would remain and which information is still missing.

Discuss your business case

Explore your process with the Automation Scan

Boermans Digital / privacy

Your cookie preferences

You decide whether Google Analytics may load. The website, scans and calculator also work without Analytics.

Completed fields, scan answers and calculator results are not sent to Analytics.

Which data and retention periods?

We remember your choice on this device for up to 180 days, including a refusal. The _ga and _ga_* Analytics cookies are configured for up to 180 days. Retention at Google is configured separately in the Analytics account.

Storage for website functionsFunctional

The website remembers your privacy choice. Only when you choose to switch language can scans, the guide and calculator temporarily carry answers within the same tab. Contact details are not stored in that transfer.

Contact form protectionAt request review

Google reCAPTCHA loads when you open the review of your completed request. Google then processes technical data and may use security cookies. This is separate from your Analytics choice. Contact by email remains available.

Analytics is off.