Workflow · Production

How Dexa transforms podcasts into an interactive knowledge base

The problem

Finding specific expert insights within hours of podcast audio content was like searching for a needle in a haystack, creating a significant barrier for professionals, learners, and decision-makers who needed to extract actionable knowledge efficiently.

Workflow diagram · grounded in source
1
Podcast audio ingestion
trigger
“The challenge of creating such a platform lies in processing vast amounts of audio content accurately and efficiently”
2
Audio transcription with speaker ID
ai_action
“Dexa needed a solution that could handle podcast transcription at scale while maintaining high accuracy and proper speaker identification”
3
Chapter detection and segmentation
ai_action
“the platform uses AssemblyAI's automatic chapter detection to break down lengthy podcast episodes into specific, topic-based segments”
4
User asks a question
trigger
“users can engage in what feels like a personal AMA (Ask Me Anything) session with their favorite podcast hosts, available 24/7”
5
AI surfaces expert answer with timestamp
ai_action
“When users ask questions, Dexa's AI not only provides answers, but takes them to the exact moment—or chapter—in a podcast where an expert discussed that topic”
6
Verifiable source output delivered
output
“This direct connection to source material transforms static podcasts into an interactive knowledge base where every insight is verifiable and contextual”
Reported outcome

Dexa processed millions of hours of podcast content without interruption, enabling users to ask questions and receive immediate, verifiable answers sourced directly from expert podcast conversations.

Reported metrics
Podcast content processedmillions of hours
Reported stack
AssemblyAI
Source
https://www.assemblyai.com/customers/dexa-customer-story
Read source ↗

Frequently asked questions

What did this team achieve with this AI workflow?

Dexa processed millions of hours of podcast content without interruption, enabling users to ask questions and receive immediate, verifiable answers sourced directly from expert podcast conversations.

What tools did this team use?

AssemblyAI.

What results were reported?

Podcast content processed: millions of hours (source-reported, not independently verified).

How is this workflow AI workflow structured?

Podcast audio ingestion → Audio transcription with speaker ID → Chapter detection and segmentation → User asks a question → AI surfaces expert answer with timestamp → Verifiable source output delivered.