skinny.

Latest / The CTO Podcast with Fexingo: Technical Leadership, Architecture, and Engineering Org

How Salesforce Solved the AI Hallucination Crisis

When Salesforce launched its Einstein AI capabilities, they faced a critical trust deficit: their models were generating plausible but incorrect data for enterprise clients. This episode examines how the company moved from simple accuracy checks to a rigorous 'grounding' architecture, implementing strict retrieval-augmented generation pipelines and human-in-the-loop validation. We look at the specific engineering trade-offs between latency and truthfulness, and why maintaining a single source of truth became the hardest part of scaling AI in the CRM space. The story highlights how…

Log in to save or queue this episode.

The skinny

The skinny isn't ready yet — notes appear once the transcript is processed.