Work overview 02 / AI & human judgement
Useful AI. A human in the loop.
Maintaining Intercom Fin answers, correcting exceptions and using Claude to support everyday reporting.
What I worked on
Maintaining Fin's knowledge
Replicated agent resolutions and added answers, articles and macros for recurring questions.
Reviewing exceptions
Reviewed and corrected Fin responses, including failures involving policy exceptions.
Supporting team reporting
Used Claude with Intercom, Linear and Slack connectors to draft macros, spot ticket patterns and prepare weekly support reports.
The principle behind it
AI assistance still needs judgement: the knowledge behind an answer, the exceptions it misses and the person who reviews it all matter.
Weekly reporting covers work handled, new patterns flagged and blockers for support leadership.
This overview describes responsibilities listed in my résumé, not a measured-impact case study. It does not reproduce internal workflows or customer data.
Next work overview