Enterprise Context Management for Analytics

Shane Butler

Co-founder @ AI Analyst Lab

Hai Guan

Head of Data @ Ontra | Ex-LinkedIn

+ Rachel Herrera

This course is popular

29 people enrolled last week.

Make AI analysis reliable with the right definitions, data, and business rules

🚨🚨THIS WORKSHOP IS $25, PLEASE USE PROMO CODE "CONTEXT25" DURING CHECKOUT!!!

If you keep explaining the same metrics, correcting the tables an agent uses, or investigating answers that look reasonable but are wrong, getting useful AI analysis still takes a lot of supervision. Your company may already have the relevant knowledge in documentation, semantic models, and people's heads. The challenge is making that knowledge available to the agent and helping it use the right information for each question.

In this workshop, you'll learn how to turn business definitions and data knowledge into reusable context for analytics agents. You'll practice defining metrics, organizing trusted sources, and diagnosing why an agent ignored relevant guidance. These skills help reduce repeated explanations and corrections, and give you a way to investigate mistakes when they happen.

We'll be hands-on working through authoring, governance, retrieval, and improvement in Hex. You'll use sample data to build context and test its effect on analytical answers.

You'll leave with the context you built during the exercises and a 90-day plan for applying these techniques to your actual work.

What you’ll learn

Learn to identify context problems, give agents usable business guidance, and assess the results so you spend less time correcting AI.

Recognize missing or ambiguous metric definitions, inappropriate data sources, and incorrect relationships between tables.
Distinguish an answer that is consistent from one that correctly applies your business rules, so you know what needs checking.

Express metrics, dimensions, and joins in a semantic model, and capture additional business guidance in rules and guides.
Decide where each piece of context belongs so you can reuse it across analyses and maintain it as definitions change.

Identify trusted data sources and understand how ownership, access, and governance affect which context agents should use.
Reuse definitions from existing documentation, semantic layers, and repositories, reducing duplicate work and conflicting guidance.

Investigate why an agent missed a relevant guide or semantic model, then make targeted changes to help it find and apply that information.
Understand how context should be selected for different questions, business domains, and users.

Examine conversations and context suggestions to identify recurring gaps in definitions or guidance.
Use example questions and evaluation criteria to assess whether a context change improved answers and whether it introduced new problems.

Choose an initial analytics use case and identify the definitions, data guidance, and ownership it needs.
Prioritize improvements and select questions to test, giving your team a practical starting point for implementation.

Workshop agenda

Why analytics agents need business context
A brief introduction to the errors that missing context can cause, and how definitions and data guidance help address them. We'll establish a starting point for evaluating improvements throughout.
Getting oriented in Hex
Get familiar with the workspace and the tools you'll use in the exercises. We'll connect the concepts to where context is authored, managed, and used in Hex.
Hands-on: author and govern your context
Build a semantic model and business guidance using sample data. Learn how to identify trusted sources and make decisions about ownership, access, and where context belongs.
Reuse the context your company already has
Explore how existing documentation, semantic models, and repositories can contribute to an agent's context. Learn what to reuse, what needs clarification, and how to avoid conflicts.
Hands-on: diagnose retrieval and routing problems
Investigate why an agent doesn't use the context you intended. Practice making changes to help it retrieve relevant guidance, and consider how context needs differ across users and business domains.
Observe, improve, and evaluate your context
Use conversations and context suggestions to identify improvements. Explore how the Hex CLI and Claude Code can help turn suggestions into proposed changes, and use evals to assess their effect.
Apply what you've learned to your work
Outline an initial use case for your team, the context it requires, who should maintain it, and how you'll evaluate progress. We'll close with questions and next steps.