Guide · November 2026
How to find processes to automate, and what to automate first
Start from evidence, not memory
Most automation lists start from interviews and workshops. These describe how people remember a process. They miss the exceptions, the workarounds and the small tasks that nobody thinks to mention.
A better list starts from evidence of how the work is really done. Three sources give it:
- Desktop recordings of the work, from task mining.
- Event logs from business systems, from process mining. The two are compared in task mining vs process mining.
- The records the work already leaves: tickets, cases, forms and their outcomes.
What makes a process a good candidate
Score each recurring task on these questions:
- Volume: how often is the task done, and by how many people?
- Time: how long does one case take, from start to end?
- Rules: do clear rules decide the steps, or does each case need a new judgment?
- Inputs: are the inputs digital and in a stable format?
- Variants: do most cases follow a few paths, or does every case differ?
- Outcome: can you check a result against what a person did on the same case?
- Risk: what happens if an automation gets one case wrong?
Multiply volume by time to get the hours at stake. Then use rules, inputs and variants to judge how hard the task is to automate.
What to automate first
Start with a task that has many cases, clear rules, a low cost of error and an outcome you can check. Leave rare tasks and high-judgment tasks for later, even when each case takes a long time.
The checkable outcome matters most for the first automation. If you can compare the automation’s answer with what your team did on the same past case, you can measure it before anyone depends on it.
Prove it before it goes live
Test each automation on real past cases it has never seen, away from live systems. Compare its result with what the employee did. Let it go live only when the results match and your team approves.
At Fella Health, Subliminal mapped the work, identified what to automate, and tested agents on real recorded tickets in a simulated environment, away from live systems. The full account: the Fella Health case study.
How Subliminal finds processes to automate
Subliminal records how work is really done across your teams and applications, without interviews, and maps every recurring workflow: its steps, tools, inputs, outputs and exceptions. From that map it finds the automation opportunities with the most impact, then builds and tests an automation for each one.