
The Joe Reis Show Why 90% of Your Data is Wasted (and How AI Reclaims It) w/ Amit Prakash
Apr 2, 2026
Amit Prakash, co-founder of ThoughtSpot and founder of Ampup who led AI teams at Google and Microsoft, tackles how we lose 90% of data by forcing reality into rigid tables. He discusses extracting high-fidelity signals from unstructured sales conversations. Topics include dynamic ontologies, sales as athletic strategy, AI agents and coaching, and building a central Business Brain to capture hidden value.
AI Snips
Chapters
Books
Transcript
Episode notes
Dynamic Ontologies Replace Fixed CRMs
- Dynamic ontologies are automated, context-specific schemas that change by company, stage, and product instead of fixed CRM fields.
- Amit explains they detect axes like preparation and confidence from full meeting transcripts to create these ontologies.
Measure Stage Specific Sales Drivers
- Treat each sales stage as a different objective and measure stage-specific drivers like discovery early and negotiation later.
- Use transcripts to shrink conversational words into task-relevant dimensions and then run aggregation/driver analysis.
Agents Could Run The Early Sales Dance
- Early sales is a guarded information dance where buyers and sellers reveal just enough; agents could automate that matching phase.
- Amit suggests buyer and seller agents could handle the early matching, then handoff to humans for high-trust deals.



