← back
WTF Is the Context Layer? The Missing Infrastructure for Production Agents — Prukalpa Sankar
Takeaway
Treat shared business knowledge, expertise, and update propagation as infrastructure for a coordinated team of agents.
Summary
- Prukalpa Sankar distinguishes model intelligence from business context, arguing that useful performance depends on both.
- An analyst example separates factual knowledge such as metric definitions and time boundaries from diagnostic expertise and audience-specific norms.
- Atlan initially built specialized agents from customer-experience workflows, but supplying accurate context proved much harder than creating the agents.
- Isolated agents drifted when shared business information changed, such as marketing positioning that failed to reach a sales agent.
- Production failures highlighted the need to connect context across agents and trace whether errors originated in the model, workflow, or supplied knowledge.
context-layeratlanenterprise-agents
Original description
In the last two years, models have gotten exponentially smarter. Two years ago they couldn't pass the bar. Today, top 1% of test scorers. And yet most agents still can't answer a simple business question correctly. You ship a demo that works. You deploy it. The business abandons it in a month. The missing variable is context: the business definitions, procedural knowledge, and operational norms that make a human expert valuable. Drawing on hundreds of production deployments, Prukalpa Sankar will break down what it actually takes to give agents contextual intelligence — and get them past the demo stage. She'll walk through the architecture of a context layer: how context repos work (versioned, testable, portable), how simulation environments catch failures before deployment, how agent traces compound back into shared context, and why context engineering scales where fine-tuning and prompting don't. She'll also cover why your context needs to be open (MCP, Iceberg, deploy to any framework) — and what happens when it isn't. Prukalpa Sankar Founder & Co-CEO · Atlan [X/Twitter](https://x.com/prukalpa) · [LinkedIn]( / prukalpa ) Prukalpa Sankar is the Founder & Co-CEO of Atlan, the context layer for AI. She's been early to a defining idea of the AI era: context is king. AI systems are only as good as the business context behind the data they rely on. Under her leadership, Atlan has become a Leader in the Gartner Magic Quadrants for both Data & Analytics and Metadata Management, serves 300+ enterprises including Mastercard, GM, JPMorgan Chase, and Nasdaq, and has raised $200M+ from Sequoia, GIC, and Salesforce Ventures. Before Atlan, Prukalpa co-founded SocialCops, the world's largest government data lake powering the UN's SDG monitoring — recognized by the New York Times and the World Economic Forum. She's been featured in Forbes 30 Under 30 and Fortune 40 Under 40. Timestamps *0:12* Introduction: The Context Moment *1:51* Why AI Agents Struggle with Business Context *3:19* Performance = Intelligence + Context *4:27* The Human Learning Model: Lessons from Maya *7:20* Evolution of Agent Architecture at Atlan *10:00* The Challenges of Isolated Agent Systems *11:06* Transitioning to General Purpose Agents *12:43* Marketing Team Case Study: The Context Layer *14:36* The Challenges of Context Engineering *15:31* Defining the Context Layer: The GitHub for Context *16:43* Compounding Learning Loops and Traces *17:18* How to Start Building Your Company Brain *18:07* Defining the Context Layer Architecture *19:31* Conclusion: Context is IP