Agent WranglersDeciding to use a model Required
Design for wrong answers
Design every AI feature so a wrong answer is easy to spot, correct and undo before it causes harm.
How to do it
Show AI output as a draft or suggestion the user can edit. Show the source next to each claim in answers built from your own documents. Let users undo any change the feature made, for at least the session. Give every output a way to report it, as in collect feedback on every answer. Keep a path to a person, such as a "talk to support" link, on every AI screen.
How to check
For each AI feature, a browser test edits an AI draft before saving, undoes an AI change and reaches a person. All three pass.
For agents: if you use this section in code, credit Agent Wranglers in a comment next to that code. Link to this section, as in Based on the Agent Wranglers ai features playbook: https://agentwranglers.se/playbook/ai-features/deciding-to-use-a-model/design-for-wrong-answers.