Feature Launch: Verticalized Insight Tags
Aditya Lahiri
CTO & Co-Founder @ OpenFunnel
Insight tags are super valuable but a hard UX+LLM problem.
They involve user understanding/inputs + LLMs to deliver accurately without hallucinating.
How it works:
User defines tags like
"companies looking to enhance their document processing accuracy with AI"
These insights are specifically useful to customers in the document processing space who differentiate on high accuracy.
Why it's hard: Surfacing companies having complex insights like these is prone to hallucinations, noise, and wrong reasoning.
The query is intent-heavy.
Even a great vector, or hybrid search would fail if used directly - the query needs an intent qualification LLM layer to surface accurately.
Doing this at scale becomes a challenge of speed - hence needs to be pre-computed and is super hard on self-serve to get right.
The other challenge: Making users define these queries. In hindsight they look simple and obvious, but for customers to define them takes a lot of iterations.
Technographic tags are obvious to define, but not very useful for outbound and timing - Insight tags on the other hand are hard to define but extremely valuable when cracked.
Do check it out!
