Do analytics agents need a semantic layer? Roundtable + Q&A
This is a practitioner-led panel discussion around whether semantic layers are necessary for conversational analytics. I wanted to bring together a group of data practitioners who actually implemented conversational analytics to chime in on the discussion - what do these agents need to be reliable?
The panelists have all built a semantic or context layer in production. To balance out the discussion, I even recruited someone who doesn't believe in it! 😁
Andre Baaij, previously Head of Data at Babylist
Dylan Morrish, Analytics Engineer at Graniterock
Evan Mullins, Analytics Engineer at Weedmaps
Jonathan Paquette, Data Scientist at Juul Labs
Topics/Questions:
What is a semantic layer exactly?
Did you build a semantic layer? If so, how? What tools?
What was most important in ensuring the agent doesn't hallucinate? What was most effective in providing context?
What would be the most important takeaway from your experience?
This is a virtual roundtable/panel discussion, where there will be a Q&A session after — see you there!
Your host: JetBrains Databao is a service for implementing conversational analytics. We just released a CLI that generates a semantic layer and benchmark from your dbt project. And nope, none of the panelists used Databao because we're new!