Customer Insights Agent
The answer was already in your warehouse.
Now it is just a question away.
A conversational analyst inside Gemini Enterprise. Ask in plain English, and the agent works out which of your BigQuery tables can answer it, queries them where they already live, and explains what it found.
Illustrative representation using sample data. The actual product, including its interface and functionality, may differ.
How it works
Four steps, none of which involve a ticket.
- 01
Ask in plain language
A business user types the question the way they would ask a colleague.
- 02
It works out where to look
Schemas, table locations and descriptions tell it which datasets can answer.
- 03
It queries your BigQuery
Against the datasets you approved, where your data already lives.
- 04
You get a business answer
Not a table dump — the finding, the number behind it, and what to check next.
Demo
See how it works.
A full walkthrough, from the question to the answer, inside Gemini Enterprise.
Recorded in product. No mockups.
What comes back
An answer written for a decision.
Paid search retained 41.8% of Q3 cohorts at day 90 — ahead of every other channel.
- 1The finding, in plain termsA direct answer phrased for a decision maker, not a data model.
- 2The number behind itThe figure that supports the claim, with what it actually measures.
- 3The trend that explains itA chart where a picture makes the point faster than a sentence.
- 4Where to look nextFollow-up questions worth asking, so one answer leads to the next.
Architecture
What sits where.
Delivered through Gemini Enterprise and hosted by Syntasa. Nothing is installed in your project — your users reach it, and it reaches back into the datasets you approve.
Your environment
Yours- Your usersWorking in Gemini Enterprise, where the agent sits alongside everything else.
- Your BigQuery datasetsThe ones you approve, and only those. Your rows never move.
- Your access controlsExisting warehouse governance still decides who sees what.
Syntasa’s Gemini Enterprise project
Ours- The agent, on Cloud RunConfigured, maintained and monitored by Syntasa. Nothing for you to run.
- A metadata layer in BigQuerySchemas and table descriptions — structure, not records.
- FirestoreSession state, so follow-up questions know what came before.
- Cloud KMSEncryption for credentials and connection details.
Registration and entitlement are configured at purchase, so only authorised organisations and users can reach the service.
Governance
Exactly what moves, and what doesn’t.
Stays in your BigQuery
- Every record, every row. Queries run in place — zero-copy, no ingestion.
- Any dataset you have not approved. The agent cannot reach it.
- Your access controls, deciding who sees what, exactly as they do now.
No shadow copy of your data is ever made.
Held by Syntasa
- Structural metadata — table locations, schemas, descriptions. Not your rows.
- Session and conversation state, so a follow-up has context.
- Connection credentials, encrypted with Cloud KMS.
The metadata layer is why answers are accurate without anyone hand-modelling relationships.
Every query is transparent and auditable. No black box.
Getting started
Three steps, and none of them are an implementation project.
Purchase through Gemini Enterprise
Access is bought on the marketplace listing, which connects your organisation to the agent.
Onboarding runs itself
Client registration and entitlement management are configured automatically at purchase.
Approve your datasets
Point the agent at the BigQuery datasets it may query. Existing tables onboard as they are.
The data was always yours.
Now the answers are too.
Runs against your own data, where it already lives. Pricing scales with usage — ask us.