Onboarding Team Members
When you onboard new users into an existing Compass organization, the goal is not to position Compass as perfect. The goal is to help people use it well.
Compass is strongest when users treat it like a fast, collaborative data partner:
- Great for open-ended questions and directional analysis
- Helpful for finding the right tables, metrics, and follow-up questions
- Improving over time as your team corrects it and adds context
For high-stakes decisions, teams should still review important results and keep refining Compass's context.
Recommended admin rollout
Keep the first rollout simple:
- Start with one team or one channel instead of the whole company.
- Share 3-5 example prompts that are relevant to that team's work.
- Explain what Compass is good at and where users should double-check results.
- Tell users they should correct Compass when something is off, unclear, or out of date.
- Make sure at least one admin or data team owner reviews context requests regularly.
What new users should know
Before inviting a broader group, make sure new users understand these basics:
How to ask
- Ask questions in plain language and include business context when it matters.
- Use threads for follow-up questions, exploration, and directional reads.
How corrections work
- If an answer looks wrong, incomplete, or based on the wrong definition, reply with the correction directly in the conversation.
- Corrections are not automatically checked in as approved context. They create context update requests for admins and data owners, often the data team, to review and merge or reject.
When to verify
- Important numbers should still be sanity-checked, especially for executive reporting or irreversible decisions. Use See all steps to inspect the analysis path and generated SQL behind an answer.
Kickoff message template
Use this as a starting point for Slack, email, or an internal wiki. It will be more useful if you add instructions tailored to your business and data context, such as which datasets are available, which teams or domains Compass supports, and where users should be extra careful about verification.
Compass is a fast way to explore our data in natural language. It is especially useful for open-ended questions, directional analysis, and finding the right follow-up questions quickly.
In this channel, Compass can answer questions about [available data or domains]. If you are not sure where to start, ask "What data is available?" before asking a more specific question.
It will not be perfect every time. If Compass gets something wrong, please reply in the thread and correct it. Corrections are not checked in automatically. They become context update requests that our admins and data team review, then merge or reject. Once approved, Compass will use that context in future answers.
For important decisions, please verify key numbers the same way you would with any other analytics workflow. Some domains, such as finance or RevOps, require especially high accuracy, so be extra mindful when reviewing those answers. Use See all steps to inspect the analysis path and generated SQL behind an answer.
Rollout checklist
Use this checklist before inviting a new team, opening up a new channel, or encouraging a broader group to start using Compass. The goal is to make the first experience specific enough that users know what to ask, how to verify important answers, and where corrections go.
Before launch
- Pick the first channels or teams to onboard.
- Confirm which data sources and domains are available in each channel.
- Assign an admin or data owner to review context update requests.
- Prepare 3-5 example prompts that are relevant to the team's work.
At launch
- Share the kickoff message template with business-specific details filled in.
- Explain Compass's strengths and limits up front.
- Tell users to ask "What data is available?" if they are unsure where to start.
- Tell users how to correct answers in-thread and how those corrections are reviewed.
After launch
- Review and merge or reject context requests on a regular cadence.
- Use the home page to track which datasets are queried most and which users are using Compass most often.
- Watch for repeated questions, confusing definitions, or high-value corrections that should become approved context.
- Use adoption patterns to decide where to improve context, add examples, or coordinate with teams that are using Compass heavily.
- Refresh examples or channel guidance when new datasets are added.