Intel / video

Paving the Road to Production: Governed Enterprise AI Agents — Harish Gaggar, Credit Karma

Why it matters

Shows how to route every tool call through an authenticated gateway with tracing, approval gates, and , giving a concrete reference design for production agent platforms.

Key takeaways · AI-distilled

Agents at Credit Karma are registered through an internal portal that provisions identity, memory, traces, evaluations and secret handling, then move through container builds, review, security scanning and Kubernetes deployment.
In the LangGraph workflow, threads, runs and checkpoints preserve state, while separate tracking tables record model use, tool calls, costs and evaluation fields. Sensitive actions pass policy checks and human approval.
Gaggar's caveats: PII filtering is a control to evaluate, not a guarantee, and the roughly 90% token cut from pruning comes from a demonstrated query without a paired answer-quality study or measured ROI.

Terms in this piece · Glossary

MCP — The Model Context Protocol — an open standard that lets any AI assistant plug into any tool or data source without custom integration code.
context window — The maximum amount of text a model can consider at once — its working memory for the current conversation or task.
AI agent — An AI system that doesn't just answer once but works toward a goal in a loop — taking actions, reading the results, and deciding what to do next.
context compaction — Summarizing an agent's earlier conversation to free room in the context window so a long session can keep going.