Start with frequency and volume

A task that takes five minutes but happens 60 times a week can matter more than a two-hour task that happens quarterly. Count occurrences and total time before prioritising.

Look for stable inputs and rules

Good candidates have inputs that are available in a predictable form and steps that can be described without a long conversation.

If staff regularly reinterpret the goal, negotiate exceptions or repair missing information, the task may need process design before automation.

Measure the full cost

Include not only hands-on time but also chasing, waiting, context switching, corrections and customer delay. This makes the opportunity more realistic.

Check consequence and reversibility

A low-risk reminder is different from an automated pricing decision or customer communication. The higher the consequence, the stronger the validation, approval and recovery controls should be.

Choose the smallest useful slice

Automate a bounded repeatable step first. Keep a human owner, logs and a way to recover. Measure time saved and error rate against the baseline before expanding.

Rank candidates on value and readiness

A simple two-axis view helps. One axis is value: time consumed, customer delay, error cost or owner interruption. The other is readiness: stable steps, clean inputs, known exceptions and low enough consequence.

High-value, high-readiness work is a strong starting point. High-value but low-readiness work usually needs process redesign first. Low-value work may not justify implementation and maintenance at all.

Account for maintenance from day one

Every automation becomes a small operating asset. Someone needs to know what it does, which credentials and systems it relies on, and how to update it when a form, API, staff role or business rule changes.

Include that future maintenance in the decision. A workflow that saves 30 minutes a month but needs specialist repair every quarter is unlikely to be a good trade.

Use automation to create capacity, not just activity

Time saved should have a destination. Decide whether the recovered capacity is meant to improve response, reduce overtime, increase throughput, improve quality or simply give the owner more uninterrupted time.

Without a destination, businesses often automate a task and then fill the recovered time with another layer of administration.