Automation
What should a small team automate first?
A decision framework for choosing automation work that saves time without creating a fragile AI project.
The best first automation is rarely the most impressive demo. It is a frequent, measurable workflow with predictable inputs, an obvious owner and a safe fallback.
01
Look for delay and repetition
Start with work that happens many times a week: copying lead details, assigning enquiries, preparing a standard response, moving information between systems or finding the same internal policy.
Estimate the current time cost and the cost of delay. A five-minute task performed twice a month is not a priority. A lead that waits six hours for routing may be.
02
Prefer stable rules before generative judgement
If a workflow can be expressed as clear rules, automate those rules first. Use generative AI for summarising, classifying or drafting where its output can be checked—not as a silent decision-maker for high-risk outcomes.
Define what happens when data is missing, a tool is unavailable or confidence is low. The fallback is part of the product, not an edge case to add later.
03
Run the smallest useful pilot
Choose one team, one workflow and one measurable outcome. Keep a human approval point while you observe exceptions. Log what the system received, what it produced and what the person changed.
A useful pilot proves reliability and operational fit. It also reveals whether the real problem was process design, data quality or access—not a lack of AI.
04
Know when not to automate
Avoid workflows with unclear ownership, constantly changing rules, poor source data or consequences the business is not prepared to review. Automating disorder usually makes it faster and harder to see.
A small manual process with good visibility can be the right answer. The goal is better operations, not an automation count.