Start with the cost of current work
Measure how often the task occurs, how long it takes and how many people touch it. Add rework, errors and waiting where relevant.
A five-minute task performed hundreds of times can deserve more attention than a three-hour task that happens once a quarter.
- Frequency
- Minutes per run
- People involved
- Approximate cost
- Errors or delay
Check whether the process is stable
Automation needs rules. If the process changes every week, relies on unstructured conversations or contains too many unknown exceptions, redesign may come first.
Stable does not mean rigid. It means inputs, major decisions and outputs can be described with enough precision.
- Known inputs
- Repeatable rules
- Identifiable exceptions
- Verifiable output
Prioritise impact against effort
A simple impact-versus-effort view separates quick wins from structural projects. Strong first candidates usually combine volume, low ambiguity and reasonable access to the systems involved.
Do not ignore risk. Automating an irreversible or regulated action needs more control than automating draft preparation.
- High volume + clear rules
- Digital data
- Available APIs/integration
- Low consequence of error
- Easy-to-measure result
Design control and measurement from day one
Automation should leave a trace: what it received, what it did and which exception it found. Without visibility, saved effort can turn into difficult investigations.
Set a metric before launch: time saved, cycle time reduced, fewer errors or released capacity.
- Logs
- States
- Exception alerts
- Human review where needed
- Outcome metric
Put numbers behind manual work.
Use the free Ordexian calculator to estimate the annual hours and direct effort cost of a repetitive task.
Open calculator →