AI without hallucinations: verified answers instead of AI guesses.
oneAgent uses AI to understand questions in natural language. The metrics themselves, however, follow defined data sources, business rules, filters, time references, and verification — so teams get answers that are traceable, reproducible, and safe to use.
Test environment with prepared shop data · no credit card · no data connection required
What was net revenue per customer in 2024?
| Müller GmbH | €128,400 |
| Schmidt KG | €96,700 |
| Becker AG | €81,250 |
Plausible answers aren't enough when decisions depend on numbers.
Generic AI can summarize text, develop ideas, and explain patterns. For company metrics, however, more is needed: access to the right data, consistent definitions, valid filters, role permissions, and verified calculation logic.
Numbers can sound plausible — and still be wrong
A language model generates answers probabilistically. Without access to your defined data logic, it cannot reliably know which table, which time period, which cancellation logic, or which filter applies to a metric.
Business terms aren't unambiguous
"Revenue" can mean gross revenue, net revenue, invoiced revenue, paid revenue, or revenue after returns. Without a defined metric, the same question produces different answers.
The problem isn't AI itself. The problem is AI without verified business logic.
The AI understands the question. The metric stays controlled.
oneAgent separates language understanding from metric calculation. The AI recognizes the question, time period, context, and terms. The calculation follows an approved metric with a defined source, logic, filters, and verification status.
AI for language
User question, natural chat, and AI language understanding. Business teams can ask questions like in ChatGPT.
Business logic for numbers
Data source, metric, filter, time reference, business rule, and validation. The calculation follows approved logic — not freely generated guesses.
A controlled layer between AI, data, and business logic.
oneAgent combines AI language understanding with semantic mapping, business rules, and validation. Chat feels flexible — without metrics being freely interpreted.
The AI doesn't freely decide which number is correct. It guides users to the correct, approved logic.
Defined once. Verified and approved. Reused in chat.
For every important metric, it's defined what exactly is measured, which source applies, which filters are used, and who is responsible. Once a metric is verified and approved, oneAgent uses the same logic in chat — in three traceable steps.
Every important metric starts with a clear definition.
oneAgent makes business terms machine-readable and unambiguous. So "revenue" doesn't mean something different depending on the user, report, or tool — it points to a defined metric with clear logic.
Source
Which table, view, API, or data source is authoritative?
Calculation
Which aggregation, column, or formula is used?
Time reference
Order date, invoice date, payment date, or another time reference?
Filters & business rules
Which filters are required, which are optional, and which rules always apply?
Aliases
Which terms point to the same metric — revenue, turnover, sales, monthly revenue?
Owner
Who is the business owner who can review or approve changes?
Before a metric appears in chat, it gets verified.
Key metrics go through verification before they're used as trustworthy. That turns a definition into an approved metric that can be used reproducibly in chat.
Test cases
Test cases are set up for typical questions and expected results. This checks whether the metric behaves correctly in relevant situations.
Comparison against the source system
Results are compared against the authoritative system or existing reference reports. This shows whether the calculation matches the business truth in the company.
SQL validation
The generated query is technically validated before being used as the basis for answers. Faulty or invalid queries shouldn't silently end up in chat answers.
Business approval
Only once the definition, tests, and comparison check out is the metric approved as trusted.
Trust isn't created by a claim. Trust is created by definition, verification, approval, and transparency.
Every answer shows which logic was used.
For company metrics, the number alone isn't enough. Users need to see which metric was used, which time period applies, which filters are active, and what logic produced the result.
Metric
Which approved metric was used?
Time period and time reference
Which time period was queried and which date field applies to the analysis?
Filter
Which required filters and user filters were applied?
Business rule
Which business logic applies — cancellation logic, period logic, or channel definition?
Source / details
Which data source or detail view can be checked?
This lets business teams work quickly without blindly trusting a black box. And BI, data, or controlling teams can check the logic behind an answer.
New chatoneAgent·Yesterday·12:43 PMControlled AI use instead of uncontrolled data exports.
oneAgent is designed for use with company data: with a role and permission model, traceable data sources, verified metrics, and deployment options for different security requirements.
Roles and permissions
Users only see the data, metrics, and areas they're authorized for.
No shadow AI workarounds
Teams can generate analyses, summaries, and ideas within a controlled environment instead of copying data into external AI tools.
Traceable data sources
Answers show which metric and which logic were used. This keeps it visible where results come from.
Deployment options
oneAgent can be operated in the cloud, on your own infrastructure, or on-premise, depending on requirements.
GDPR-compliant architecture
Data access, the role model, and infrastructure are planned so that company and data protection requirements are met.
Hosting in Frankfurt and operation in a Microsoft Azure environment are possible depending on setup.
Metrics need ownership — not just an answer in chat.
oneAgent makes visible who owns a metric, what logic is behind it, what verification has taken place, and when a metric was approved.
Owner
Every central metric can be assigned to a business owner or team.
Approval status
Users can see whether a metric has been verified and approved as trusted.
Controlled changeability
When a metric is changed, the new logic must be verified and approved again in a traceable way.
Verification history
Tests, comparisons, and validation steps can be made visible as part of the governance process.
Consistent definitions
Teams work with the same metrics instead of parallel spreadsheet definitions or conflicting reports.
Verified metrics, AI analysis, and forecasts are clearly distinguished.
oneAgent shouldn't suggest that every interpretation or forecast is a verified actual metric. That's why it must be clear whether an answer is based on an approved metric, an AI analysis, or a forecast.
Verified actuals
Revenue, margin, return rate, contribution margin, inventory. These metrics follow defined sources, calculations, filters, and verification.
AI analysis & summaries
Explaining patterns, writing management summaries, formulating possible causes, or developing campaign ideas. This content is based on verified data but is interpretive AI support.
ML forecasts
Revenue forecast for the next 12 months, demand forecast by category, or a country forecast. Forecasts are predictions based on historical data and shouldn't be presented as verified actuals.
Trust also comes from the fact that oneAgent doesn't label everything the same way.
More on how oneAgent is used and positioned.
Use cases by team
How leadership, controlling, sales, e-commerce, IT, and other teams use oneAgent day to day.
See use cases →In comparisononeAgent vs. BI, Excel & generic AI
Where oneAgent is strong and where other tools fit — control, flexibility, and AI use compared directly.
See the comparison →For your data warehouseVerified answers on your data warehouse
How oneAgent makes SQL Server, Snowflake, Databricks, and more usable via chat — live-connected, with deterministic calculation instead of AI guessing.
See the data warehouse solution →Frequently asked questions about verified answers and security
Answers about metric logic, validation, roles, security, and data connections.
oneAgent uses AI to understand questions and phrase answers clearly. For verified metrics, however, the calculation follows defined data sources, business rules, filters, and verification. That's why we speak of verified answers rather than freely generated AI guesses. We deliberately avoid blanket claims like "never wrong" or "100% correct".
A verified metric is defined from a business perspective and technically validated. This includes source, calculation, time reference, filters, aliases, and owners. Test cases, comparison with the source system, and SQL validation can also be applied before the metric is approved as trusted.
This happens together with the responsible teams, for example controlling, finance, BI, data, or the relevant business team. oneAgent shouldn't automatically guess your business logic — it should make defined logic usable.
Answers can show which metric was used, which time period and time reference apply, which filters are active, and whether the metric is verified. That makes the answer not just fast, but traceable.
AI analyses, summaries, or ideas can be based on verified data but are interpretive AI support. They should be labeled differently than verified actual metrics.
No. Forecasts are predictions based on historical data and ML methods. They can build on verified historical data but are not guaranteed predictions and should be labeled accordingly.
oneAgent is built for controlled use with company data: with a role and permission model, traceable data sources, verified metrics, and cloud, own-infrastructure, or on-premise options depending on requirements.
No. oneAgent is meant to complement existing BI, controlling, and data governance and make it usable for business teams. The logic isn't replaced — it's made accessible in chat.
Yes. The free trial runs for 14 days with prepared shop data, predefined metrics, and guided use cases. Your own data is currently connected together as part of a supported pilot project.
See for yourself how verified answers happen in chat.
Try oneAgent for free with prepared shop data — or book a demo if you have questions about metric logic, validation, roles, security, or data connections.
With prepared shop data · no credit card · no data connection required. See how it works
Not sure what fits? Compare trial, demo & pilot