Surya, good to read this, and I get the initial hesitation. We had the same reaction on our side a year ago.
What changed the picture for us wasn't Genie getting smarter on its own, it was what we put around it. Once we added clear instructions per space, exposed a few Unity Catalog SQL functions as trusted assets for the calculations that can't be wrong (revenue, churn, anything finance signs off), and ran benchmarks before opening it to business users, the accuracy went from "impressive demo" to something we can defend in a steering committee. Agent mode with multi-step reasoning helped a lot with the messier questions too.
I'm curious about the telecom case though. How are they using it for the data science pipeline itself?
Thanks for sharing.
Thomaz A. Rossito Neto
Principal Data & AI โ CI&T
thomazn@ciandt.com
linkedin.com/in/thomaz-antonio-rossito-neto