Every team has its own version of the numbers.
Sales, finance and operations each calculate the same KPI their own way. Meetings start with whose number is right, and simple questions wait days for the data team.
- Two dashboards, two numbers for the same KPI
- KPI logic buried in SQL and spreadsheets
- Simple questions queued for analysts
- AI chatbots that guess what a KPI means
Define, answer, check with review.
We turn each KPI into one governed definition, agreed with finance and the business owner.
- One definition per KPI
- Versioned in Git
- An owner for every metric
People ask in plain language where they already work. The agent answers from the governed definitions.
- Number and explanation
- Query and definition shown
- Your access rights apply
It can build dashboards from the same definitions and flag tiles that drift from them, in BI tools such as Power BI, Tableau or Looker.
- Dashboards from governed KPIs
- Mismatches flagged
- Fixes approved by an analyst
What teams ask.
- “Which deals slipped out of this quarter?”
- “How did win rate change by region?”
- “Which accounts are ordering less than usual?”
- “Why is gross margin down this month?”
- “How does DSO compare with last quarter?”
- “Which cost lines grew fastest this year?”
- “Which warehouses missed their delivery target last week?”
- “Where are returns rising?”
- “How long does an order take from payment to shipping?”
A semantic layer comes first.
AI Analyst answers from a semantic layer: the governed definitions of your KPIs, with their joins and filters, kept in code. Without one, any AI model has to guess what a KPI means from table names, and two people asking the same question can get different numbers. If you don’t have a semantic layer yet, we build it as the first step.
dbt Semantic Layer, Snowflake semantic views, Databricks metric views, Cube or LookML. We use the one that fits your stack.
Read-only access to your warehouse or lakehouse, set up so each query runs with the asker’s own permissions.
Slack, Microsoft Teams or your BI tool. Every answer names the definition and the query it used.
For each KPI, agreed with finance and versioned in Git.
Every answer shows the query and the definition behind it.
The agent never changes a KPI definition or dashboard on its own. It proposes each change, and one of your analysts approves it.
Questions we hear
How is this different from a chatbot on our data?+
A generic chatbot writes SQL against raw tables and guesses what a KPI means. AI Analyst queries only the definitions you approved, and shows the query and definition with every answer.
We already have Cortex Analyst or Genie. Why AI Analyst?+
Use them. Where they fit, we build AI Analyst on Snowflake Cortex Analyst or Databricks AI/BI Genie, and add what they need to give correct answers: the semantic layer, a test set of real questions, and access from Slack or Teams. If your data sits on more than one platform, or your dashboards are in Power BI, Tableau or Looker, we add our own agent layer on top.
Does it change dashboards on its own?+
No. It proposes each change, and an analyst reviews and approves it.
Which tools does it work with?+
It connects to your warehouse or lakehouse, your BI tool and the channels your teams already use, such as Slack or Microsoft Teams. We confirm your stack during scoping.
Where does the model run?+
On the AI service already in your cloud account: Amazon Bedrock, Snowflake Cortex, Databricks Mosaic AI Model Serving or Vertex AI. On-premises, we host open-source models on your infrastructure.
Who can see which numbers?+
We set it up so each query runs with the asker’s own access rights, so people only see data they are already allowed to see.
Can we add it to a running platform?+
Yes. It can be part of a Production Sprint or added to a platform that’s already live.
Where does it run, and who controls it?+
Everything we build for you is yours: pipelines, data models, infrastructure code and documentation, in your own cloud account and repositories. Our agents and accelerators come with a licence that keeps working even if you stop working with us. If you need full source access, we offer that too.
Stop debating whose number is right.
Tell us which platform and BI tool you run.