AIRA classifies a call by comparing its transcript against the description of each meeting type configured in your workspace. The matching only works as well as those descriptions are written.
In plain English: after a call ends and the transcript is ready, AIRA reads through your full list of configured meeting types and picks whichever one's description best matches what was actually discussed on the call.
Worked example: Recruiterflow's built-in Intake type is described as: "Working session with a hiring manager or client to define the search. Covers ideal profile, must-haves, compensation band, team context, and hiring timeline." If a recruiter has a call with a client where they walk through exactly that, AIRA matches the transcript to that description and tags the call Intake. A different call, where a candidate is checked in on post-placement with no evaluation involved, matches the Check-in description instead: "A touchpoint with a candidate already in pipeline or post-placement. No evaluation — just relationship maintenance, settling-in pulse, or keeping the door open for future roles."
A few things that follow directly from this:
If two of your meeting types have similar or overlapping descriptions, AIRA is more likely to mix them up. Distinct, specific descriptions produce more reliable tags.
A meeting type with a blank or vague description gives AIRA very little to match against, so calls that should map to it may get tagged as something else instead.
Tagging happens after the call is fully processed, not live during the call.
This only runs at all if "Use AIRA to auto-classify meeting type" is turned on in Notetaker → General Settings.
See Writing Effective Meeting Type Descriptions for guidance on getting the most reliable results.