AI on Your Data Warehouse — Verified, Not Guessed
oneAgent extends your existing data warehouse with a governed AI access layer for business teams. Users ask questions in natural language and get live results from approved data — based on defined metrics, business rules, and governed SQL queries.
Try it for free with prepared demo data. Connecting your own data warehouse happens during joint onboarding.
New chatoneAgent·Today·09:12Your data warehouse → oneAgent → verified chat answer
How a question becomes a governed answer.
The AI interprets the language. Data access, metrics, business rules, and query logic are defined by IT, BI, and the business.
From business question to live result
- 01
Business question
A business team asks a question in natural language.
- 02
Language understanding
The AI identifies the intent, time period, and relevant terms in the question.
- 03
Approved metrics and rules
The semantic layer determines which metrics, tables, relationships, and business rules may be used.
- 04
Governed SQL query
Within these boundaries, oneAgent generates an executable SQL query against the data warehouse.
- 05
Live result with provenance
The result comes from the current query and shows the data source and logic used.
The AI interprets the question — it does not define what revenue, margin, or region mean. Which data is used and how a metric is calculated stays governed.
IT, BI, and the business govern these points together
Data access scope
Which data sources, roles, and areas are permitted for the query.
Tables and relationships
Which tables, joins, and data relationships apply to the use case.
Metric definitions
How metrics are calculated, named, and approved for use.
Business rules
Which filters, time periods, and business rules are applied to the query.
Security
How permissions, confidential data, and permitted use are protected.
Three questions business teams can soon answer themselves.
oneAgent delivers tables, visualizations, and traceable results directly from the existing data warehouse.
“How has revenue by region developed over the last quarter?”
Answer
Table and chart showing revenue development by region.
Basis
Approved revenue metric and agreed regional mapping.
Traceability
The data source, time period, and filters used are shown.
“Where are we off plan, and where is the gap biggest?”
Answer
Plan-versus-actual variance by area, sorted by the size of the gap.
Basis
The responsible controlling team's agreed plan-versus-actual calculation.
Traceability
The data source, time period, and calculation logic used are shown.
“How has a metric developed over the last 12 months, and what's next?”
Answer
Metric time series with a forecast for subsequent periods — clearly labeled as a forecast, not a verified actual.
Basis
The defined time series and agreed forecast model.
Traceability
Forecast inputs and assumptions are visible and agreed with the responsible team.
Runs where your data already lives
oneAgent connects directly to your existing data warehouse — without an additional analytics warehouse and without recurring data exports.
SQL Server
Direct connection to on-premise or Azure-hosted SQL Server instances.
PostgreSQL
Connects to managed or self-hosted PostgreSQL databases.
Snowflake
Access to your existing Snowflake warehouses through your current account structure.
Databricks
Direct queries against Databricks SQL warehouses and Delta tables.
Oracle
Connection to Oracle databases within your existing operating model.
Microsoft Fabric
Connects to the Fabric Warehouse and SQL endpoints of lakehouses.
Operating model — two alternatives
Managed by oneLake
Operated in a dedicated Azure environment in Frankfurt.
Self-hosted
Operated in your own cloud or data center environment.
- Security concept defined jointly with your team
- Existing roles, permissions, and security concepts are respected and technically enforced during the connection
Native AI features are strong — within their own ecosystem. Why oneAgent?
If your data lives entirely within one platform, that platform's native AI can be a good fit. oneAgent is built for organizations that want one governed chat layer across existing data warehouse systems and operating models.
Native platform solution
- Deep integration into that one ecosystem
- Uses the governance and infrastructure already in place there
- Especially useful with a homogeneous platform strategy
- Features, cost, and operations are tied to that one vendor
e.g. Databricks Genie, Snowflake Cortex, Microsoft Fabric Data Agent
oneAgent
Cross-platform- One interface across different data warehouse platforms
- Central metrics and business rules
- One consistent rollout and governance approach
- Managed or self-hosted operation
- No need to switch your existing data platform
You don't replace your data warehouse. You add a shared, governed access layer on top of it.
See detailed comparisons →Is oneAgent the right next step?
A pilot works especially well once a concrete business need meets a solid data foundation.
Good prerequisites for a pilot
- Your data warehouse is in productive use and the relevant data is available.
- A defined user group has recurring questions that existing reports can't answer flexibly enough.
- A clearly scoped use case can be defined for the pilot.
- Business, BI, and IT jointly validate metrics, rules, and results.
This should be clarified before the pilot
- iThe focus is on document search or knowledge management rather than structured business data.
- iUsers, questions, and success criteria for the pilot are not yet defined.
- iNo one owns the relevant metrics and data yet.
- iThe required data sources are not yet integrated or aligned on the business side.
If some prerequisites are still missing, we'll jointly clarify in an architecture call which steps make sense before a pilot.
From use case to a production pilot.
Five steps to validated answers from your data warehouse.
- 01
Define the use case and success criteria
We define the users, specific business questions, and the criteria used to evaluate the pilot.
- 02
Review data access and scope
We review the platform, network, permissions, and the tables required for the use case.
- 03
Configure metrics and rules
Together we define relationships, metrics, business rules, and security requirements.
- 04
Validate the answers
Business and BI teams review reference questions, query logic, and results.
- 05
Activate pilot users
After sign-off, the agreed users start using the pilot.
Pilot outcome
A connected use case, defined metrics and rules, validated reference questions, and a solid basis for deciding on a wider rollout.
Frequently Asked Questions
Answers to technical questions about operations, cost, and data connectivity.
In the live-only model — without importing data — you pay EUR 25 per user per month, with no additional storage costs. If you also import selected data, storage costs apply proportionally. Self-hosting is negotiated individually.
oneAgent only lets the AI understand the question linguistically. The answer itself is produced by a fixed layer of approved metrics and business rules that generates deterministic SQL, executed live against your data warehouse. The AI never guesses a number — it is calculated and checked against the agreed definition.
No. In live mode, oneAgent connects directly to your data warehouse. The query runs live against the existing infrastructure, so no second copy of your data is created.
Not necessarily. oneAgent can complement existing BI, DWH, and database structures and make them usable as a flexible self-service AI layer on top.
Power BI Copilot is tied to the Power BI platform and works primarily on pre-built Power BI data models. oneAgent, by contrast, is data-warehouse-agnostic and connects directly to your data warehouse — SQL Server, PostgreSQL, Snowflake, Databricks, Oracle, or Microsoft Fabric — regardless of whether you use Power BI. The SQL query is calculated deterministically, not guessed by the AI.
Yes. oneAgent is container-based and runs either hosted in Frankfurt or on-premise in your own data center, right next to your data warehouse.
SQL Server, PostgreSQL, Snowflake, Databricks, Oracle, and Microsoft Fabric. For Fabric customers, oneAgent complements the existing Fabric warehouse with an additional chat layer for business teams.
oneAgent respects your existing roles and permissions model instead of bypassing it. Who can see which metrics and data sources is defined together with IT and BI.
Yes. oneAgent runs either hosted in Frankfurt (Azure) or on-premise in your own data center — your data is not exported uncontrolled into external AI tools. Your existing roles and permissions model stays fully in place.
Does oneAgent fit your existing data architecture?
In an architecture call, we review your use case, data platform, security requirements, and a possible pilot scope.
Not sure what fits? See trial, demo & pilot at a glance →