Building Agentic Apps With Craft: Field Stories from Austin Vance, Co-Founder or Focused
Convergence PodcastMay 01, 202500:56:49

Building Agentic Apps With Craft: Field Stories from Austin Vance, Co-Founder or Focused

What does it actually take to build agentic AI applications that hold up in the real world? In this episode, Ashok sits down with Austin, founder of Focused, to share field stories and hard-won lessons from building AI systems that go beyond flashy demos. From legal assistants to government transparency tools, Austin breaks down the concrete criteria for identifying where AI makes sense — and where it doesn’t.

They unpack how to find the right starting point for your first agentic app, why integration with legacy systems is the real hurdle, and the engineering must-haves that keep AI behavior safe and reliable. You’ll hear practical guidance on designing eval frameworks, using abstraction layers like LangChain, and how observability can shape your development roadmap just like in traditional software. Whether you’re a product leader or a CTO, this conversation will help you distinguish hype from real opportunity in AI.

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...
A practical checklist for identifying your first AI-powered app
The hidden cost of "AI for AI’s sake" and where traditional software is better
Why repetitive knowledge work is prime territory for automation
How Focused helped Hamlet build an AI for parsing government meeting data
Where read-only data access gives you a safe starting point
Why integration is often more complex than the AI itself
The importance of eval frameworks and test-driven LLM development
How to use observability to continuously improve AI agent behavior
Speed vs. believability: surprising lessons from Groq-powered inference
Using multiple models in one system and LLMs to QA each other

Mentioned in this episode
Hamlet (government transparency startup) - https://www.myhamlet.com/?convergence
LangChain - https://www.langchain.com/?convergence
Groq - https://groq.com/?convergence
Claude (Anthropic) - https://claud.ai/?convergence
Dspy Prompting framework - https://dspy.ai/?convergence
Shopify AI memo (referenced) - https://convergence.fm/episode/shopifys-leaked-ai-mandate-explained-6-takeaways-for-your-product-team?convergence
Amazon Bedrock / SageMaker - https://aws.amazon.com/bedrock/?convergence

Subscribe to the Convergence podcast wherever you get podcasts including video episodes to get updated on the other crucial conversations that we'll post 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.  

Follow the Pod
Linkedin: https://www.linkedin.com/company/convergence-podcast/
X: https://twitter.com/podconvergence
Instagram: @podconvergence