Ep220: The AI Architecture We Deleted - featuring Demandbase
Notes de l’épisode
Demandbase's Vice President of Product explains why they deleted a fully-approved AI architecture two months before launch — and how the rebuild surpassed some of their customer’s highest expectations.
Topics Include:
- AWS's Achint Naveen introduces Demandbase's VP of Product, Chad Holdorf
- Demandbase unifies sales, marketing, and revenue data into one view
- November's architecture used many specialized agents, all committee-approved
- That design failed constantly — only a 30% conversation pass rate
- On May 11th, the team deleted the entire architecture
- Rebuilt in May with AWS Strands: one simpler, flexible agent
- Pass rate leapt from 30% to 94% almost overnight
- Week two retention rose from the low 20s to upper 80s
- Weekly active users grew 45% week-over-week after launch
- Real customer interviews play, calling the new AI a "dream"
- One user cut an hour-long report down to fifteen minutes
- Customers now trace ad impressions directly to closed deals
- Holdorf's advice: delete and rebuild when architecture gets too complex
- AWS's Naveen walks through Bedrock, AgentCore, and Strands
- AgentCore Memory highlighted as Demandbase's next area of exploration
Participants:
- Chad Holdorf – Vice President of Product Management, Demandbase
- Achint Naveen – Sr Account Manager, Amazon Web Services
See how Amazon Web Services gives you the freedom to migrate, innovate, and scale your software company at https://aws.amazon.com/isv/