In short: Microsoft's Fabric Data Agent is a new AI tool within the Fabric ecosystem that translates natural language questions into SQL, DAX, or KQL queries. Promising in theory — but it only works in English, requires cross-geo AI processing that may move your data outside your Azure region, and demands a complete Fabric infrastructure. oneAgent offers native multilingual support (German + English), GDPR-compliant processing in Frankfurt, and 550+ native connectors — starting at 25 EUR per user per month.
What is the Microsoft Fabric Data Agent?
The Microsoft Fabric Data Agent is Microsoft's latest approach to AI-powered data analysis — and it's distinctly different from the well-known Power BI Copilot. While Copilot works within Power BI reports, generating DAX formulas and suggesting visualizations, the Data Agent is a standalone, configurable artifact within Microsoft Fabric.
In practice: you create a Data Agent, connect it to up to five data sources from your OneLake, and give it instructions on how to answer questions. The agent translates natural language questions into SQL, DAX, or KQL queries and returns results as text or tables — no visualizations.
The Data Agent has been generally available since March 2026, making it a very young product. Several features remain in preview. Microsoft positions it as a tool for analysts and business users who need quick answers from their Fabric data.
Important distinction: The Data Agent is not Power BI Copilot. Copilot works within existing Power BI reports; the Data Agent queries your OneLake data sources directly. Both require Fabric capacity, and both require your data to reside in Microsoft's OneLake.
For a comparison with Power BI Copilot specifically, see our dedicated article: oneAgent vs. Power BI Copilot.
The Fundamental Difference: Microsoft Ecosystem vs. Open Platform
The core of this comparison fits in one sentence: The Fabric Data Agent analyzes data that already lives in Microsoft's OneLake. oneAgent connects directly to your existing systems.
This sounds like a small technical detail, but it has far-reaching consequences:
Fabric: Everything Must Go to OneLake First
Before the Data Agent can answer a single question, your data must be in Microsoft Fabric. That means building data pipelines, creating semantic models, defining governance rules. For companies using Shopware, SAP, DATEV, or industry-specific systems, this is a 2-to-6-month project — requiring expertise in Spark, Python, Delta Lake, DAX, KQL, and Azure administration.
oneAgent: Direct Connection
oneAgent connects via 550+ native connectors directly to your existing systems. Shopware, SAP, Microsoft Dynamics, DATEV, Salesforce, PostgreSQL, REST APIs — no intermediate layer, no data migration. Typical setup takes days, not months.
What This Means for Mid-Market Companies
Mid-market companies rarely have a data team that can build and maintain Fabric pipelines. They have an ERP, a CRM, maybe an e-commerce platform — and they want answers from that data. oneAgent meets this reality; the Fabric Data Agent assumes a different one.
What Does Microsoft Fabric Cost vs. oneAgent?
Microsoft Fabric's pricing is capacity-based and notoriously difficult to estimate from the outside. The Data Agent has no separate price — it runs on your Fabric capacity and consumes Capacity Units (CUs).
Fabric Capacity Pricing (Pay-as-you-go, as of April 2026)
| SKU | CUs | Price (PAYG) | Price (Reserved, 1 Year) |
|---|---|---|---|
| F2 | 2 | ~$263/month | ~$175/month |
| F4 | 4 | ~$526/month | ~$350/month |
| F8 | 8 | ~$1,051/month | ~$701/month |
| F16 | 16 | ~$2,102/month | ~$1,401/month |
| F32 | 32 | ~$4,205/month | ~$2,501/month |
| F64 | 64 | ~$8,410/month | ~$4,900/month |
Prices approximate. Actual costs depend on Azure region.
Additionally required: Power BI Pro licenses for each user (~$14/user/month) or Fabric licenses.
Realistic Scenario: 50 Users
| Cost Item | Fabric Data Agent | oneAgent |
|---|---|---|
| Fabric capacity (F32 Reserved) | ~$2,501/month | — |
| User licenses (50 x Pro) | ~$700/month | — |
| oneAgent licenses (50 users) | — | €1,250/month (~$1,375) |
| Azure overhead (storage, networking) | ~$100–300/month | — |
| Monthly total | ~$3,301–$3,501 | ~$1,375 |
| Annual total | ~$39,600–$42,000 | ~$16,500 |
| Onboarding / setup | $15,000–50,000 (consulting) | from €1,500 (~$1,650) one-time |
Factor 2.5x on running costs — and factor 10–30x on initial setup. Plus: The F32 capacity is shared across all Fabric workloads. When the Data Agent is under load, it competes with your other pipelines.
Cost Comparison: 20 Users
| Cost Item | Fabric Data Agent | oneAgent |
|---|---|---|
| Fabric capacity (F8 Reserved) | ~$701/month | — |
| User licenses (20 x Pro) | ~$280/month | — |
| oneAgent licenses (20 users) | — | €500/month (~$550) |
| Monthly total | ~$981+ | ~$550 |
| Annual total | ~$11,770+ | ~$6,600 |
Note: F8 is the absolute minimum for production use. For intensive usage or larger datasets, F16 or F32 is more realistic.
Feature Comparison in Detail
| Feature | oneAgent | Fabric Data Agent |
|---|---|---|
| Natural language queries | Yes — chat interface | Yes — chat interface |
| Supported languages | German, English, and more | English only |
| Data sources per agent | Unlimited (550+ connectors) | Max 5 sources per agent |
| Recommended tables | No limit | ~25 tables recommended |
| Max result size | Configurable | 25 rows x 25 columns |
| Visualizations | Automatic charts and graphs | None — text and tables only |
| AI architecture | Deterministic AI layer | LLM-based (probabilistic) |
| Hallucination risk | Eliminated by deterministic approach | Present (known reliability issues) |
| Response time | 2–8 seconds | 10–30 seconds typical |
| Unstructured data | Via connectors (documents, PDFs) | Not supported |
| Data processing location | Frankfurt (EU) | Cross-geo AI processing — data may leave your Azure region |
| On-premise option | Yes | No (cloud-only) |
| GDPR compliance | Full — Frankfurt hosting, on-premise available | Limited — cross-geo processing is problematic |
| Setup time | Days (plug-and-play) | 2–6 months (Fabric infrastructure required) |
| Required expertise | None (chat interface) | Spark, Python, DAX, KQL, Azure admin |
| Vendor lock-in | Data source-independent | Microsoft ecosystem (OneLake only) |
| Product maturity | Production since 2024 | GA since March 2026 |
| Minimum cost/month | €25 (1 user) | ~$981+ (F8 + 1 Pro license) |
| 14-day free trial | Yes — no credit card | No (Fabric capacity required) |
Where Is the Fabric Data Agent the Better Choice?
We believe in honest comparisons. Here are scenarios where the Fabric Data Agent makes sense:
1. You Already Run a Productive Fabric Environment
If your organization already uses Microsoft Fabric for data engineering, data science, or real-time analytics and the capacity costs are already justified, the Data Agent is a logical add-on. Your data is already in OneLake, governance is established — the marginal effort is low.
2. Your Team Works in English
Since the Data Agent only understands English, it works well for international teams or companies where English is the working language. For German-speaking business departments, it's a barrier.
3. You Need Deep Integration with Fabric Workloads
If you orchestrate Fabric notebooks, Spark jobs, and data pipelines, and the Data Agent is part of a larger analysis workflow, the native integration offers advantages that an external tool can't replicate.
4. Enterprise with a Dedicated Data Team
Large organizations with a data team that builds and maintains Fabric pipelines can position the Data Agent as a self-service layer for business departments. The complexity sits with the data team, not the end users.
5. You're Strategically Betting on Microsoft's AI Direction
Microsoft is investing heavily in Fabric and Copilot. If you're strategically committed to the Microsoft ecosystem and view the current limitations (language, cross-geo, reliability) as temporary, early adoption could make sense.
Where Is oneAgent the Better Choice?
1. Your Team Works in German (or Any Non-English Language)
This is the most obvious point — and the most important one. The Fabric Data Agent supports English only. From Microsoft's documentation: "The Fabric data agent doesn't currently support non-English languages." For a mid-market company in the DACH region, where sales, procurement, or management operates in German, this is a dealbreaker.
2. Data Privacy and GDPR Are Business-Critical
The Fabric Data Agent requires cross-geo AI processing — meaning your data may leave the Azure region where your Fabric capacity is provisioned for AI processing. For companies in regulated industries (finance, healthcare, public sector) or with strict data protection policies, this is a significant concern.
oneAgent processes all data in Frankfurt — no exceptions. A full on-premise installation is available as an option.
3. You Want to Be Productive Quickly
The Fabric Data Agent requires your data to be in OneLake. If you currently use Shopware, SAP, or an industry-specific ERP, you first need to build data pipelines. That takes 2 to 6 months and requires expertise in Spark, Python, and Azure.
oneAgent connects directly to your existing systems. Typical setup: a few days.
4. You Need Reliable Results
oneAgent's deterministic AI layer translates questions into exact database queries. The result is always correct and traceable. The Fabric Data Agent uses a probabilistic language model — Microsoft itself documents that the agent sometimes ignores custom instructions and loses conversation history.
5. Your Data Doesn't (Only) Live in Microsoft
With over 550 native connectors, oneAgent connects your Shopware data with your SAP, your CRM, and your accounting system — without an intermediate layer. The Fabric Data Agent can only query OneLake data. Everything else must be migrated first.
6. You Want Insights, Not Just Tables
oneAgent delivers automatic visualizations — charts, graphs, trend lines. The Fabric Data Agent outputs only text and tables, limited to 25 rows and 25 columns per response.
Who Should Use Which Tool? A Clear Decision Guide
| Criterion | oneAgent | Fabric Data Agent |
|---|---|---|
| Working language | German and/or English | English only |
| Company size | 10–500 employees | 500+ with a data team |
| IT budget for analytics | From €6,000/year (~$6,600) | From €11,000/year (realistically €35,000+) |
| Existing infrastructure | Diverse systems (SAP, Shopware, CRM, ERP) | Microsoft Fabric already in use |
| BI expertise in-house | Low to medium | High (Spark, DAX, KQL, Azure) |
| Primary users | Business departments, management | Analysts, data team |
| Data privacy requirements | High (regulated industries, on-premise) | Standard (cross-geo acceptable) |
| Time-to-value | Days | Months |
| Product maturity expectations | Mature and stable | Early adopter (GA since 03/2026) |
Quick Formula
- Mid-market without Fabric: oneAgent. The language barrier alone is decisive.
- Enterprise with productive Fabric environment: Fabric Data Agent as a complementary self-service layer — but evaluate the cross-geo issue.
- Mid-market with Microsoft 365 (no Fabric): oneAgent. You'd need to build an entire Fabric infrastructure first.
- Regulated industries: oneAgent — for Frankfurt hosting, on-premise option, and deterministic results.
- International team, English as working language, Fabric in place: Fabric Data Agent can be a good fit.
Frequently Asked Questions
What's the difference between the Fabric Data Agent and Power BI Copilot?
Power BI Copilot works within existing Power BI reports: it generates DAX formulas, suggests visualizations, and summarizes report content. The Fabric Data Agent is a standalone artifact that queries OneLake data sources directly — independent of Power BI reports. Both require Fabric capacity, but the Data Agent is more flexible in data connectivity and doesn't require existing reports. For a detailed comparison with Power BI Copilot, see: oneAgent vs. Power BI Copilot.
Will the Fabric Data Agent support German soon?
As of April 2026, Microsoft has not published a roadmap for multilingual support in the Fabric Data Agent. The official documentation explicitly states: "The Fabric data agent doesn't currently support non-English languages." When or whether German support will be added remains unclear.
What does "cross-geo AI processing" mean in practice?
When the Fabric Data Agent answers a question, your data is sent to an AI service that doesn't necessarily run in the same Azure region as your Fabric capacity. This means: even if your Fabric instance is in West Europe (Netherlands) or Germany West Central (Frankfurt), your data may be transferred to another region for processing. For companies with strict data protection requirements, this is a significant risk.
Can oneAgent work with Microsoft data sources?
Yes. oneAgent natively connects to Microsoft SQL Server, Azure SQL, Microsoft Dynamics 365, SharePoint, and other Microsoft services — alongside 550+ other connectors. You don't have to choose between Microsoft and oneAgent; you can query your Microsoft data through oneAgent without running a Fabric capacity.
How reliable is the Fabric Data Agent?
As a GA product since March 2026, the Fabric Data Agent is still young. In practice, users report cases where the agent ignores custom instructions or loses conversation history. Several features remain in preview. oneAgent has been in production since 2024 and uses a deterministic AI layer that eliminates hallucinations.
What if I want to try the Fabric Data Agent?
You need an active Microsoft Fabric capacity (minimum F2 at ~$263/month) and Power BI Pro licenses. There is no free trial for the Data Agent itself. With oneAgent, you can start a 14-day free trial — no upfront costs, no credit card required.
Can I use both tools in parallel?
In principle, yes. Some organizations use Fabric for their data team and oneAgent as a self-service layer for business departments. The tools aren't mutually exclusive. In practice, though, the question is: is it worth maintaining two AI analytics platforms — or does one cover both use cases?
Next Step
Want to see how oneAgent works with your data sources — in your language, in seconds, without a Fabric setup? Book a personal demo. We'll show you the platform live with your real data. No obligations, 30 minutes.
More comparisons:
- oneAgent vs. Power BI Copilot — Detailed cost comparison and feature analysis
- oneAgent vs. ThoughtSpot — Enterprise BI comparison
This comparison was researched in April 2026. The Microsoft Fabric Data Agent has been generally available since March 2026 and is actively evolving. Prices and features may change. We update this article regularly. Last update: April 2026.