Drop in for a virtual coffee, Wednesdays at 5:30pm

Know when to step in

Decide in advance what AI handles and what comes to you. Escalation and human-in-the-loop are a design choice, not an afterthought. Name the triggers that hand it back to a person.

Say it another way

Human-in-the-loop just means deciding ahead of time which actions the AI may take on its own and which must stop and wait for you to approve. You set that line on purpose, rather than discovering it after the AI has already done something you would not have allowed.

yesnoAI takes an actionWithin the boundsyou set?AI handles itHand it to you
This is the Know when to step in box from the diagram at the top, opened up. Every action the AI takes is checked against the bounds you set. Inside them it runs on its own. Outside them it stops and hands control to you.

Doing this with Claude Code: Know when to step in

Knowing when to step in earns its keep in four ways: escalate on the real triggers, honour a handover at once, do not escalate on mood and ask rather than guess on ambiguity.

Escalate on the real triggers, not on difficulty. Hand back when there is an explicit request to do so, when the situation falls into a gap the rules do not cover, or when it cannot make progress. A hard case it can still work through is just work, not a reason to escalate.

Honour an explicit handover at once. When someone asks for a person, hand off immediately without another round of questions and pass on the context already gathered, so they pick up warm rather than cold.

Do not escalate on mood or self-confidence. An annoyed tone, or the model's own reported confidence, is a poor proxy for whether a case really needs a human. Acknowledge frustration and still offer to resolve it, escalating only if they insist.

On ambiguity, ask rather than guess. When a lookup returns several matches, ask for another identifier instead of picking one, because acting on the wrong record is far worse than one more question.

Common questions

  1. What should actually trigger an escalation?
    An explicit request to hand back, a situation that falls into a gap the rules do not cover, or being unable to make progress. A hard case it can still work through is just work, not a reason to escalate.
  2. Is frustration or low confidence a good signal to escalate?
    No. An annoyed tone, or the model's own reported confidence, is a poor proxy for whether a case really needs a human. Acknowledge the frustration and still offer to resolve it, escalating only if they insist.
  3. What should happen when a lookup is ambiguous?
    Ask rather than guess. When a lookup returns several matches, ask for another identifier instead of picking one, because acting on the wrong record is far worse than asking one more question.
  4. Someone has asked for a person. What now?
    Hand off immediately, without another round of questions, and pass on the context already gathered. That way the person picks the conversation up warm rather than starting it cold.

Back to For developers. The craft around it is Working with AI.