How we filled our calendar with meetings at Web Summit in Lisbon using custom-built AI agents.
The screenshot shows my calendar at Web Summit 2024. This was my third time at Web Summit (or fourth, if you count their Collision in Toronto), and for the first time I wasn’t bored or wandering around aimlessly. On the contrary — I was running from meeting to meeting because my calendar was pre-filled with relevant and quality meetings.
For context: Web Summit is a massive industry event with around 70,000 participants. Despite the official mobile app, booking meetings there is genuinely hard: filters are limited, all the important info is buried in bios (reading each one is exhausting), participants are flooded with spam and often miss messages, and the recommendation system is just for show.
What We Did to Solve This
1. Parsing the Participant Database
There’s no official API, of course. Had to get creative and carefully sniff the traffic to export the participant database.
2. Data Normalization
Exported everything into our own database, cleaned up and normalized the data.
3. Basic ICP Filters
With an already prepared ICP (Ideal Customer Profile), we immediately filtered out irrelevant participants by geography, role, and other explicit criteria (why burn tokens unnecessarily).
4. AI Agents for Data Enrichment
I wrote and ran several AI agents that enriched prospect information:
- Found company websites
- Found LinkedIn profiles
- Scraped important company info from websites
- Using a detailed and lengthy prompt (with specifics about our projects and solutions), extracted key fields for each prospect:
- Spheres — which of our areas align with their challenges
- Solutions — which of our solutions fit the client
- Qualification Reasoning — clear justification for why this is a match
Stack: TypeScript, OpenAI API (structured outputs), SERP API, custom parser, SQLite, Google Sheets API.
5. Manual Personalization and Lead Selection
We deliberately didn’t automate 100%. After AI enrichment, we manually reviewed the hottest and most relevant prospects and wrote personalized messages by hand. The leads remaining after all this processing were pure gold.
Results
We pulled this off in a week before the event, in super-accelerated mode. And this approach really worked great!
A huge credit for the success also goes to our SDR, who smartly sent out requests based on the enriched database, crafted excellent personalized messages, and actively helped organize the meetings themselves. Humans and robots working together.