Complete your Databricks User Groups profile!

Fill out a few details about yourself so the community can get to know you.
Bangladesh Databricks User Group

100,000 Agents. 1 Quadrillion Tokens. One Platform.

Summary: Kazi Tanvir discusses insights shared by Databricks at the Data + AI Summit 2026, emphasizing that while the core agent loop is a small part of building agents, the majority of the work lies in token capacity management, secure deployment, evaluation, context management, governance, and sharing. Highlighting Agent Bricks, a platform designed to address these challenges, Kazi Tanvir notes its widespread adoption, with over 100,000 agents and significant data processing capabilities. The unique feature is its integration of data and AI, facilitating comprehensive governance and analysis. Kazi Tanvir invites others to share their biggest challenges in moving agents to production.
AI Summary

The Agent Loop Is the Easy Part.

Everyone's building agents. But here's what Databricks admitted at Data + AI Summit 2026: the core agent loop is only 1% of the work.

The other 99%? The hidden cost nobody talks about:
🔸Token capacity management
🔸Secure deployment
🔸Evaluation and monitoring
🔸Context and memory
🔸Governance and sharing

That's exactly what Agent Bricks is built to solve.

Since launch, over 100,000 agents have been built on the platform, now processing more than 1 quadrillion tokens per year. Customers like AstraZeneca, Fox Corporation, and Block have already shipped production agents on it.

What makes it unique: it connects data and AI in one place. Because agents don't just consume data, they produce it. Every reasoning trace, action, and memory output needs to be governed and analyzed.

For data engineers and architects in Bangladesh building AI systems, this is the architecture shift to watch. The agent runtime is becoming part of the data platform.

What's your biggest challenge when moving agents to production? 👇

1 comment