Subvolts take Learn why AI projects fail and when not to use large language models. Sep 29, 2025 - Developers
- Learn practical lessons on building and evaluating RAG pipelines and integrating reasoning models, and understand when LLMs are not the best fit.
- Business
- Understand common reasons AI pilots fail and gain insights into effective evaluation and error analysis for AI projects.
This episode provides practical lessons from a year of building AI systems, offering valuable advice for practitioners.
We've all done RAG, now what?
Audio streams from Daniel Whitenack and Chris Benson. Your place is remembered on this device.
From Subvolts
Summary and analysis by Subvolts editorial, AI-assisted.
Rajiv Shah returns to Practical AI to discuss lessons learned from a year of building retrieval-augmented generation (RAG) pipelines and integrating reasoning models. Hosts Daniel Whitenack and Chris Benson explore why AI pilots often fail, the importance of evaluation and error analysis, and when large language models are not the right solution for enterprise challenges.
This episode talks about building AI systems that help find information. It explains why many of these systems don't work well and why checking for mistakes is important. It also covers when using a powerful AI language tool is not the best choice for a business problem.
This episode offers practical insights for AI practitioners and business leaders. It addresses common pitfalls in deploying AI solutions, emphasizing essential data science skills like evaluation and error analysis. The discussion on when LLMs are not the answer is particularly relevant as frontier labs like Google Gemini and OpenAI continue to push LLM capabilities, while companies need to balance innovation with practical application. It's for anyone building or considering AI projects in an enterprise setting.
- HostDaniel Whitenack
- HostChris Benson
- GuestRajiv Shah
- Guest: Rajiv Shah
- Title: We've all done RAG, now what?
- Show: Practical AI
- Length: 44 minutes
- Topics Covered: Lessons from building RAG pipelines, reasoning model integrations, why AI pilots stumble, evaluation and error analysis, when LLMs are not needed.
Chapters from the publisher ↗ · Transcript ↗ Key moments and transcript search come with a later piece.
Longtime friend of the show Rajiv Shah returns to unpack lessons from a year of building retrieval-augmented generation (RAG) pipelines and reasoning models integrations. We dive into why so many AI pilots stumble, why evaluation and error analysis remain essential data science skills, and why not every enterprise challenge calls for a large language model.
Featuring:
Upcoming Events:
- Join us at the Midwest AI Summit on November 13 in Indianapolis to hear world-class speakers share how they’ve scaled AI solutions. Don’t miss the AI Engineering Lounge, where you can sit down with experts for hands-on guidance. Reserve your spot today!
- Register for upcoming webinars here!
- RAG
- Rajiv Shah
- Retrieval-Augmented Generation: A way to help AI models find and use specific information to answer questions better.
- retrieval-augmented generation
- Rajiv Shah
- A method that helps AI find and use specific information to improve its answers.
- reasoning models
- Rajiv Shah
- AI systems designed to think through problems and make logical connections.
44 minNarrative intelligence and the human advantage
WithReed Frerichs
47 minOpen models and the future of Physical AI with NVIDIA
WithMing-Yu Liu
49 minFrom AGENTS.md to Enterprise Deployment
WithNick Kuhn
Questions people ask
- What is the episode about?
- The episode unpacks lessons from a year of building retrieval-augmented generation (RAG) pipelines and reasoning model integrations, discussing why AI pilots stumble and when LLMs are not the best solution.
- Who is Rajiv Shah?
- Rajiv Shah is a longtime friend of the show who returns to discuss practical AI development and integration challenges.
Page generated Oct 19, 2026. Summaries are Subvolts' own; the audio, artwork and notes belong to Daniel Whitenack and Chris Benson.