← All tools

Enterprise Context Management for Analytics

30-Day Plan

WEEK 1

Make your analysis reusable with another member of your team

Two teammates comparing an analysis

You know why you trust an answer after working through it. See whether someone else on your data team can use what you learned without you explaining every step.

What to do this week

  1. Invite one data teammate into your analytics workspace, so they can ask questions using the same data.
  2. Give them a question you’ve already worked on. Have them ask it in a new thread before showing them your answer.
  3. Compare the answers, sources, and definitions. Add any important rule you had to explain aloud to your shared context, then try again.
  4. Switch roles. Let them bring a question they know better and see whether you can use the context they contribute.

How to do this in Hex

Ask a workspace admin to invite your teammate from Settings > Users > + User and confirm they can use the same data connection. Each person should open a new Thread with the same question. Open the Notebook behind each Thread to check the queries and calculations. Add a broad rule to workspace context or a subject-specific rule to a Guide, then use Test and Publish.

You can try Hex here. Invite teammates · Threads · Workspace context and Guides

WEEK 2

Make your analysis more actionable with a decision maker at your company

Colleagues discussing a chart together

Your data teammate can help check how an answer was calculated. The person using it can tell you what decision they need to make and whether you answered the question they meant to ask.

What to do this week

  1. Invite at least one person in product, marketing, sales, finance, or another team who would use an answer you worked on in Week 1 into your analytics workspace.
  2. Ask what decision they are making, what they mean by the important terms, and what could make the answer misleading.
  3. Have them ask the question in their own words. You’ll find they ask questions quite differently than how the data team might. Review the answer together, as well as how the agent got there, update or add the relevant context, and try it again.

How to do this in Hex

Invite the decision maker as a workspace member if they need to ask their own questions. Have them start a Thread and share it with you. Open the Notebook behind the answer together, then use Context Studio to update and test any guidance that was missing.

You can try Hex here. Sharing and permissions · Threads · Guides

WEEK 3

Connect your AI analyst to knowledge across the company

Colleagues sharing documents and data

Some of what your teammates explained is already written down in company documents, code, data definitions, or past analyses. Connect one of those sources so you can use the original work and keep track of where the knowledge came from.

What to do this week

  1. Choose one company system that holds knowledge relevant to your questions. It might be Notion, Linear, Google Drive, Confluence, a code repository, or something else. Connect it to your analytics workspace if an integration is available, as we did with Notion in the workshop.
  2. Ask the agent to find the relevant document, analysis, or definition. Open the original and check its owner, date, and whether it still applies.
  3. Review what you found with its owner. Add the reusable part to your shared context: company-wide context, a domain guide, a semantic model, or a description on past analyses or a data table and the columns it contains.
  4. Then ask a question that should use it. Check whether the agent found and applied it.

How to do this in Hex

In Hex, go to Context Studio > Apps or Settings > Integrations to see the available External Apps. Your workspace admin may need to enable the one you choose. Connect a relevant source as we did with Notion, then ask a Thread to find useful material. Open the original and review it with its owner before adding anything to shared context. For warehouse data, use a data connection.

You can try Hex here. External Apps · Data connections · Guides · Semantic models

WEEK 4

Prioritize the questions your company needs answered

An analyst reviewing completed work

Make a short list of the decisions your company is making now and next quarter. Use it to decide which questions and context changes deserve attention first.

What to do this week

  1. Look at your team’s roadmap, planning docs, and notes from product, sales, finance, and leadership meetings. List the decisions being made now and those coming up next quarter.
  2. For each decision, write down the questions people ask, the information they use today, and what they will need to know to make the decision well. Add questions you saw in Threads or heard from colleagues.
  3. Try the most important questions in your analytics workspace. Check the answer and how it was produced. Record what worked, what was missing, which source or definition needs attention, and who owns it.
  4. Rank the list by business impact, urgency, how often the question comes up, the cost of a wrong answer, and the effort to fix it. Choose the first one or two context changes to tackle.
  5. Share the list with the people who own the decisions and the data. Agree on an owner and reviewer for each chosen change, then put time on the calendar to build and test it. Show one question that now works better than it did in Week 1.

How to do this in Hex

As you invite more people into your shared Hex workspace, you can see more of the questions being asked across the company. In Context Studio > Observability, review topics and warnings to find recurring questions and answers that need attention. Use Thread Inspector to check the highest-priority cases from your roadmap or meetings. Add those gaps to your list, then test the first proposed change in Context Studio or with an eval suite.

You can try Hex here. Agent Observability · Evals · Guides

After 30 days

You should have shared context shaped by people from your data team and other teams, one connected source of company knowledge, and a prioritized list of questions to improve next. Share the list with the people who own those decisions and put time on the calendar for the first changes.

Questions and reading

For questions that come up as you work, browse the AI analytics FAQ. For a longer explanation of how to write and manage context, read The Weekly Query on context engineering and management.

Need help with a step?

Tell us which question you're working on and where you're stuck. You can leave out company details you cannot share.

Email AI Analyst Lab →

Want to do this with your team?

AI Analyst Lab advises companies scaling AI analytics across teams. We can help your team decide where to start, work with the people who own your definitions and data, and set up a way to test and maintain the context. Book time to discuss what that could look like for your company.

Book a conversation