In this episode, Karthik shares hard-won lessons about what it really takes to make agentic workflows work in production — and why most companies are still getting it wrong. We talk about the structural failures hiding inside most AI proof-of-concepts, the overlooked importance of eval frameworks, and why enterprise leaders may be unintentionally blocking their own GenAI progress by clinging to control instead of creating trust.
Unlock the full potential of your product team with Integral's player coaches, experts in lean, human-centered design. Visit integral.io/convergence for a free Product Success Lab workshop to gain clarity and confidence in tackling any product design or engineering challenge.
Inside the episode...
Why most AI agent workflows break outside the demo environment
How to architect agents like a real team, with roles and review loops
What eval frameworks are and how they protect optionality
Common failure patterns when swapping models or changing prompts
How to tell when you’ve actually hit product-market fit
Why pivot decisions are rarely data-first and often emotionally fraught
How startups maintain speed by designing for change, not certainty
What big companies get wrong about AI adoption and control
Lessons from a $400M trading system failure Karthik lived through
The future of agent-to-agent communication and orchestration protocols
Mentioned in this episode
DoubleO.ai
Craft.co
Ignition
Ocar.io
Ragas eval framework
CrewAI
AutoGen
LangGraph
Granola (AI call recorder)
Subscribe to the Convergence podcast wherever you get podcasts, including video episodes on YouTube at youtube.com/@ConvergenceFMPodcast
Learn something? Give us a 5-star review and like the podcast on YouTube. It’s how we grow.