Volume and time
How often the task occurs and how much human time it consumes.
Use Case Finder methodology
How AROG AI scores automation use cases
The Finder turns a short process description into three automation hypotheses. It separates what the model may suggest from what the application calculates, so the financial estimate can be inspected and changed.
The output is an initial hypothesis. It is not an offer, audit, implementation specification or guarantee of savings.
Inputs used by the Finder
- 1
The process description supplied by the user, including frequency, team size and current manual steps when available.
- 2
Suggested weekly hours removed, implementation complexity and indicative delivery time proposed by the model.
- 3
The hourly work cost and implementation cost selected in the results. These assumptions are editable after the Finder runs.
What makes a use case stronger
Cost of work
The editable loaded hourly cost used only by the deterministic calculator.
Repeatability
Whether the process follows stable steps rather than changing case by case.
Data readiness
Whether the required inputs are available, usable and owned by the business.
Integrations
Which systems must exchange data and whether reliable access exists.
Exceptions
How often the normal path fails and what a safe escalation looks like.
Human oversight
Where a person reviews, approves or can reverse the system's action.
Risk
Legal, operational and reputational impact if the output is wrong.
When not to automate
- No accountable process owner can define success or accept exceptions.
- The process changes faster than the team can document and test it.
- The system would make irreversible decisions without review or rollback.
- Legal or safety risk is high and the result cannot be independently verified.
- The available data is incomplete, unlawfully sourced or too unreliable for the intended decision.
Confidence and data limits
Low
The description is vague, volumes are missing or dependencies are unknown.
Medium
The workflow and main systems are known, but exceptions or data quality still need validation.
High
The process is stable, measured, reversible and has a named owner with test data.
Deterministic value model
The model never supplies the displayed money. The application caps each recommendation at 15 hours per week and all recommendations at 35 hours per week, then applies the same formulas everywhere.
Monthly value = capped weekly hours × 4.3 × hourly work costAnnual value = monthly value × 12Payback months = implementation cost ÷ monthly valueFirst-year ROI = (annual value − implementation cost) ÷ implementation cost × 100%Illustrative example
A team estimates that triaging repetitive enquiries could remove 10 hours of manual work per week. The example below uses the public default assumptions, not client data.
Monthly value
$1,505
Annual value
$18,060
Payback
3.3 months
First-year ROI
261%
Actual scope, integration effort and results require process validation. The example does not predict a result for another company.