Choosing which business processes to automate first

Most organisations have more automation ideas than capacity to deliver them. Choosing well at the start builds confidence and frees up time for the next project. Choosing badly, by automating something rare, unstable or high-risk, can make people sceptical of automation altogether.

Look for these characteristics

• High volume: the task happens often enough that small savings add up.
• Clear rules: most cases follow a predictable pattern.
• Stable process: the steps aren't about to change.
• Digital inputs: the information already exists in a system, form or document.
• Measurable: you can record how long it takes and how often errors occur today.

Estimate value and risk together

For each candidate, estimate the time spent per month and the cost of an error. A task that takes many hours but where mistakes are easy to spot and fix is usually a better first project than one with high stakes, such as payments or regulatory submissions, even if both save similar time.

Do you need AI, or just automation?

Rules-based automation, such as moving data between systems, sending notifications or creating records from form submissions, is predictable, cheaper to run and easier to test. AI becomes useful when the input is unstructured, for example free-text emails, varied documents or requests that need summarising or classifying.

When AI is appropriate, design for oversight: let people review outputs where errors matter, log what the system did, and provide an easy way to correct it. Review the data-handling terms of any AI provider before sending it business or personal information.

Start small and measure

1. Record a baseline: time taken, volume and error rate.
2. Automate one well-defined process, including how exceptions are handled.
3. Run it alongside the manual process briefly to confirm results.
4. Compare against the baseline and share the results with the team.
5. Use what you learned to choose the next process.

A modest, well-measured first success is more valuable than an ambitious project that's hard to evaluate. It gives you evidence, and a team that trusts the next automation.