Design automations with a clear exit for people

Automating a predictable task can reduce repeated work. Automating without defining what happens when faced with ambiguity only transfers the problem to customers or colleagues. The design needs to include situations in which the tool must stop.

List normal paths and exceptions

Map expected inputs, known errors, and conditions that require judgment. Examples: lack of information, contradictory messages, urgent request or high impact of an incorrect answer. Define which cases automation can safely resolve.

Create a readable transfer

When the flow escalates to a person, send the minimum necessary context, what has already been attempted, and why the task was escalated. Warn the user about the transfer and do not make them repeat the entire story unnecessarily.

Monitor and learn from referrals

Record volume, reasons, resolution time and failures. Review whether rules are routing too much, too little, or to the wrong team. Don't try to eliminate all referrals: some are signs that the process is protecting quality and trust.

Run a limited pilot, provide a way to ask for help, and pause the automation if errors with important consequences appear. Revise the drawing when the process or input type changes.

Frequently asked questions

When should I forward it to someone?

When there is ambiguity, lack of data, request outside the scope or significant consequence. Determine these criteria with whoever is responsible for the process.

Does the transfer need to include the entire history?

Include useful and authoritative context while minimizing unnecessary data. The person must understand what happened without receiving irrelevant information.

How do you know if the automation is working?

Evaluate correct completion, failures, total time, satisfaction and referral volume. No single metric describes quality.

To continue learning with practical ideas for business, subscribe to the Ímpetus newsletter for free.