By AI agents fail at retrieval when they can only match exact strings. Enterprise data lives in meeting notes, customer calls, support tickets, and documents where the words in a question never match the words in the source. Keyword search returns nothing. The agent stalls.
In this interview on TFiR, Jean Lafleur, Co-founder and Chief Operating Officer at Airbyte, breaks down how semantic search works, why it is the critical unlock for agentic AI, and what it looks like against real enterprise data sources.
Guest: Jean Lafleur, Co-founder and Chief Operating Officer at Airbyte
Show: TFiR
Here is what every platform engineer and AI practitioner building agentic systems needs to know.
Technical Deep Dive
Q: What is semantic search and how does it differ from keyword search for AI agents?
Jean Lafleur, Co-founder and Chief Operating Officer at Airbyte, explains that semantic search uses natural language questions to find information by topic, not by matching exact wording. Where keyword search requires the query to match the precise terms in the document, semantic search finds conceptually related content even when the words differ. This makes it fundamentally different in how it surfaces relevant information across large, varied data sets.
“Semantic search is really the biggest unlock for agents in terms of finding information on unstructured data.”
Jean Lafleur, Co-founder and Chief Operating Officer, Airbyte
Q: What are real examples of semantic search queries AI agents run against enterprise data?
Lafleur gives concrete examples: an agent can answer questions like “What customer concerns were discussed at last week’s planning meeting?” or “Where have we discussed pricing strategy across documents?” or “What are customer conversations where there were pricing concerns?” These queries operate across indexed meeting notes, call recordings, and documents. No exact keyword match is needed because the agent retrieves by topic.
“It could be, what customer concerns were discussed at last week’s planning meeting? Where have we discussed pricing strategy across documents?”
Jean Lafleur, Co-founder and Chief Operating Officer, Airbyte
Q: What types of unstructured data can semantic search index for AI agents?
Lafleur notes that semantic search applies to any indexed unstructured source: meeting notes, customer conversations, internal documents, and support records. The examples he gives span engineering issues, login failures, pricing discussions, and planning meetings, covering multiple teams and data types. Any topic, any team’s data becomes retrievable once it is indexed.
“Engineering issues, login failures, any topic, any teams could have some.”
Jean Lafleur, Co-founder and Chief Operating Officer, Airbyte
Resources & Documentation
- Airbyte, open-source data movement platform for syncing data from any source into your data stack for AI and analytics use cases
***
👇 Click to Read Full Raw Transcript
Swapnil Bhartiya: Jean, is it possible for you to share some real example of how semantic search helps customers?
Jean Lafleur: Semantic search is actually how agents think they use natural language questions. It could be, what customer concerns were discussed at last week’s planning meeting? Where have we discussed pricing strategy across documents? Or Milinos? What are customer conversations where there were pricing concerns? And so they look at the topic and find the topics across every conversation that has been dexed or every meeting notes, all of that. And so it’s very different from just search, where you just look at the exact wording, how it’s being used. And so semantic search is really the biggest unlock for agents in terms of finding information on uncharted data. And so, yeah, there’s so many examples, engineering issues, login failures, any topic, any teams could have some.
