Augmeta for Snowflake and your warehouse: find out why a KPI moved, with your data where it is
Augmeta connects to Snowflake, BigQuery and Redshift with read-only access to the schemas you choose, so an AI Operator can query your tables where they live and investigate why a KPI moved. Your warehouse stays the system of record. The Operator takes each number further: whether the movement is real, why it happened, what it is worth, and what to change, with your team setting the goals and guardrails.
Why does the warehouse matter in a KPI investigation?
Because it is where a number gets settled. Web analytics tells you what the tracking recorded. Your warehouse and your order system tell you what the business recorded. When the two disagree, that disagreement is often the finding.
Augmeta’s published case studies show it:
What can an Operator do in your warehouse?
| What | What it covers |
|---|---|
| Read the schema | The tables and columns in the schemas you connected, so it writes queries against what is really there |
| Run read-only queries | One SELECT at a time, with rows capped. Anything that is not a read is refused before it runs |
| Stay inside its scope | A query that touches a schema you did not choose is refused, whatever else the role can see |
| Track KPIs you define | An administrator can write a KPI as SQL, and the Operator watches it |
How does Augmeta connect to Snowflake?
No password or private key travels. Augmeta generates a key pair, keeps the private half, and gives your administrator a setup script with the public half.
Queries run on the warehouse you name and use its credits, so a small one is the right choice.
BigQuery connects with a Google Cloud service account that has the BigQuery Data Viewer and BigQuery Job User roles on the datasets you choose. Only organization administrators can set it up. Redshift connects through an AWS role in your own account that Augmeta assumes, with access granted on exactly the schemas you choose. Your warehouse needs no public endpoint.
Do you have to move your data?
No. Operators query your data in place during an investigation. Your tables are not exported into an Augmeta database, and query results are processed to produce the answer and then discarded. The longer version, including what is stored and how it is isolated, is in Does an AI analytics agent need access to your data warehouse?
What changes when the Operator can see more than the warehouse?
The warehouse holds the definitive number, but rarely the reason it moved. So the Operator reads it next to your analytics, your session replay, and the releases and tickets that record what shipped and when, and checks them against each other. You can ask a question that crosses systems, for example: “Compare our revenue numbers with the Jira tickets about billing issues.” The approach is described in What is Agentic KPI Ops?
What happens after the answer?
Augmeta does not stop at an insight. Inside the goals and guardrails your team sets, an Operator can propose a configuration change to remedy what it found, open a Jira issue with the bug report or improvement already written up, or coordinate the team that makes the change. Then it checks whether the fix worked. What runs on its own, what needs approval and what stays with a person is a setting, not a fixed answer.
The full posture
Frequently asked questions
Is there an AI agent for Snowflake that can explain why a KPI moved?
Yes. Once Snowflake is connected to Augmeta, an AI Operator reads the schemas you chose, runs read-only queries, and works out why a KPI moved, reading your other connected tools as well.
Does Augmeta copy our warehouse data?
No. Operators query your data in place, and query results are processed to produce the answer and then discarded.
Can the Operator change anything in Snowflake?
No. It runs read-only SELECT statements through a read-only role, and anything else is refused before it runs.
Can we limit what it can see?
Yes. You choose the schemas. A query that touches any other schema is refused, whatever else the role can see.
Does it use our Snowflake credits?
Yes. Queries run on the warehouse you name and use its credits, so we suggest a small one. The Operator prefers aggregates and its results are capped.
Which warehouses does Augmeta connect to?
Snowflake, BigQuery and Redshift.
Does Augmeta train models on our warehouse data?
No. Augmeta uses foundation models through their APIs and does not fine-tune or train models on customer data. Each organization’s data is strictly isolated.
See an Operator on a KPI from your warehouse
Pick one KPI your team reports from the warehouse, and bring it to a demo.
Related reading: Does an AI analytics agent need access to your data warehouse?, How to check an AI analytics tool’s numbers against your own reporting, Augmeta with Jira and GitHub and Augmeta security.