GABI 2026 Speaker Intent Enrichment
For every speaker on the roster: LinkedIn URL, follower counts, the theme they are pushing, and the exact hook phrases their clip should catch.
The problem
"The video team needed to know, before the conference, what each of dozens of speakers was actually talking about publicly right now, so they could spot the sound bite the moment it happened on stage. Doing that read on every speaker by hand was not going to happen."
The solution
One flow. Every input type.
The workflow reads the speaker roster from a Google Sheet (Mainstage 2026 tab) and skips any row already enriched, so re-runs are safe. One speaker at a time, it uses Serper with a site:linkedin.com/in query plus the speaker's affiliation to find the LinkedIn URL, hands that URL to an Apify HarvestAPI actor to scrape the profile (with neverError so a failed scrape does not kill the run), and runs two more Serper queries for press or podcast mentions and for X and Instagram presence. A bundle step compiles the speaker's about, recent posts, press blurbs, and detected socials into one context block, plus a LinkedIn activity tier (High, Medium, Low, None). Claude Sonnet 4 reads the bundle and returns strict JSON: a 2 to 3 line intent note, 3 to 4 exact hook phrases the speaker is likely to say, and a one-line note on their X or Instagram presence. The formatted row writes back to the same sheet, matched on speaker name, and a 2-second rate-limit wait sits between speakers.
Results
Industry
Events / Video Production
Service type
Client Automation
Timeline
1 week
Tools used