Question 1

How is this different from a data catalog?

Accepted Answer

A catalog waits for people to fill it in. The Context Agent pulls definitions from your tools, drafts what is missing, fixes what Dot gets wrong, and sends each change to your team for review. It can complement a catalog by keeping definitions current.

Question 2

What can it pull context from?

Accepted Answer

dbt (re-synced daily), Looker, Tableau, Power BI, Sigma, Metabase, Malloy and Cube; GitHub, GitLab and Bitbucket; Notion, Confluence, Jira, Slite and Google Drive; uploaded documents; and your warehouse's own comments and query history.

Question 3

Can it learn from our existing dashboards?

Accepted Answer

Yes. It reads Tableau calculations and the exact values a tile shows, Looker explores and saved queries, Power BI models and Sigma tiles. It turns them into metric definitions and trusted answers to test Dot against, and it can rebuild a Tableau, Power BI, Sigma or Metabase dashboard in Dot, checking every chart against the original.

Question 4

How do I know Dot is accurate?

Accepted Answer

Give it questions with answers you trust. Train tests Dot on each one, fixes its context in a separate environment until the answers match, and reports accuracy before and after. Evaluations rerun the same questions whenever you like, including from CI, so a change that breaks a number is caught before it reaches anyone.

Question 5

What if the Context Agent writes something wrong?

Accepted Answer

Nothing goes live on its own. Every change arrives as a proposal with a readable diff. You can test it in its own environment, merge some files and keep the rest open, or reject it with a reason, and every file keeps its history.

Question 6

What is the semantic layer?

Accepted Answer

An optional Malloy model the Context Agent writes over your warehouse: reusable measures, dimensions and joins, checked against your existing dashboards and reviewed like any other change. Dot then answers through those definitions. It supports BigQuery, Snowflake, DuckDB and Dot's managed warehouse, and existing Malloy projects in Git connect too.

Question 7

How does Dot tell me when an answer had to guess?

Accepted Answer

When an answer rests on a choice your workspace does not settle, such as an undefined metric or two notes that disagree, it is marked "Assumption made" with one sentence on what was assumed. It is not a probability. Correct it in the chat and your correction becomes a proposal for review.

Question 8

How long does setup take?

Accepted Answer

Point it at your sources and review the proposals it drafts. Most teams are running within days.

Teach Dot your business. Prove it learned.

Five jobs, one agent. You approve every change.

Pulls context from dbt, your dashboards, Git and your docs.

Learns from every conversation with your approval.

Builds a semantic layer your metrics live in.

Trains and tests itself on answers you already trust.

Every change is reviewed like code, then merged by you.

3 weeks of docs, done in 3 days.

Read how Dot helps great companies be smarter, faster

Your data, your rules.

Zero data retention

SOC 2 Type II

SSO and role-based access

GDPR compliant

Full audit logging

Questions