Asky and Microsoft Copilot
Different layers of
the same problem
Asky vs Microsoft Copilot: what each layer can actually query
Asky is a governed query layer for live operational databases: it turns a plain-language question into read-only SQL, runs it inside your own deployment under your access policies, and returns a chart with the business read. Copilot works across Microsoft 365 content and modelled Power BI or Fabric data. Asky does not replace Copilot; it can be called from it.
Keep Copilot where your employees already work. Use Asky as the controlled data layer when they need live operational data inside your own query policies and deployment boundary.
PostgreSQL, MySQL, Oracle, SQL Server and ClickHouse, queried where they already run.
None required. No semantic model to publish first, no copy of the data to keep in sync.
Docker on your own infrastructure, including air-gapped, or your cloud tenant with your keys.
Scope facts, not performance claims. Measured numbers (accuracy, response time, deployments in production) are shared with their methodology during evaluation rather than asserted here.
Which Copilot are we comparing?
“Copilot” covers several Microsoft products with different scopes. Only one of them addresses database querying, so identifying the deployed product is step one in any real comparison.
Microsoft 365 Copilot
Word, Excel, Outlook, Teams and SharePoint content via Microsoft Graph. It is not a native operational SQL layer, but connectors and agents can reach external systems, and Asky can be added as one of them.
ComplementaryPower BI / Fabric Copilot
Data modelled in Power BI semantic models or held in Fabric / OneLake. It answers only once the data has been moved and modelled.
Overlapping scopeCopilot Studio
Builds custom agents that call external tools. It reaches data only through the tools it calls, and Asky can be registered as one.
Integration pathCan Copilot query an Oracle or PostgreSQL database directly?
Compared against Copilot in the Fabric / Power BI scope, the only one that queries data. Sources: public Microsoft documentation, last verified August 2026. Asky column reflects the on-prem Docker deployment.
| Dimension | Microsoft Copilot | Asky |
|---|---|---|
| Where data resides | Microsoft cloud: OneLake, Fabric or a Power BI model | Your own database. In the local deployment nothing leaves your network; in cloud BYOK the prompt reaches your nominated model endpoint, result rows do not |
| Live production data | Answers over a published semantic model; DirectQuery can reach some sources without copying data, but the model layer is still required | Queried directly on Oracle, PostgreSQL, MySQL, SQL Server, ClickHouse |
| Preparation | Semantic model, DAX measures, capacity provisioning | Knowledge vault and metric definitions; can start table by table |
| Data movement | ETL or mirroring; a copy of the data is created | None; the query travels to the data |
| On-prem / air-gapped | Cloud-based | Full on-premises deployment via Docker |
| Data residency | Subject to Microsoft regional policy | Runs on your infrastructure; no cross-border transfer |
How are access controls enforced in each?
Enforced in the Asky query pipeline itself, not configured per report, and verifiable during a technical evaluation. Microsoft column per public documentation, last verified August 2026.
| Control | Microsoft Copilot | Asky |
|---|---|---|
| Row-level security | In Power BI, defined in the semantic model | Enforced structurally on the query tree, independent of any model |
| Column-level security | Depends on model design | Denied columns cannot be returned |
| Per-user SQL visibility | No equivalent concept | Role-based, with per-user override |
| Pre-execution safeguard | Not available | Row-count estimate before execution; queries that would overload the database are blocked |
| LLM consumption limits | Governed by licensing and capacity | Three tiers: session, user and team quotas |
| Bring your own LLM key | Not available; Microsoft-provided models | Supported; the key remains with you |
| Customer-managed keys | Within the Microsoft ecosystem | Supported via AWS, GCP or Azure key management |
| Audit trail | Via Microsoft Purview | Action, outcome, source address and applied security filters recorded |
What happens when a definition or a schema changes?
| Capability | Microsoft Copilot | Asky |
|---|---|---|
| Learning definitions | Reflects what IT has modelled | Retains user- and team-level definitions learned in conversation |
| Approving knowledge | No equivalent concept | Team conventions require explicit human approval before taking effect |
| Metric governance | Not surfaced to the user | Each metric carries its source, owner, freshness and attestation status |
| Schema change detection | The model is republished | Live schema continuously compared against documentation; differences reported |
| Hallucination safeguard | Depends on model quality | Queries referencing columns absent from the schema are rejected before execution |
What does it cost on top of a Microsoft licence?
| Dimension | Microsoft Copilot | Asky |
|---|---|---|
| Commercial model | Per-user subscription plus Fabric capacity | Annual agreement priced by business unit and connected data scope, not per seat; adding readers does not change the bill |
| Other cost factors | Capacity, data movement and storage, modelling effort | One-off implementation and documentation preparation |
| Ecosystem dependency | Tied to the Microsoft stack | Database engine and LLM provider are both interchangeable |
| LLM cost visibility | Absorbed into capacity; not visible per query | Per-query cost with team and user breakdown |
| Large result extraction | Subject to Power BI export limits | Asynchronous export at million-row scale with time-limited download links |
| Reporting | Power BI; mature and widely adopted | Panels generated from conversation, through the same governance path |
Where Copilot is the right tool
Point of use
Already embedded in Teams, Outlook, Word and Excel, with no change in user behaviour.
Document productivity
Summarizing documents, drafting correspondence and meeting notes are outside Asky’s scope by design.
Enterprise reporting
Power BI remains a mature platform for visualization and distribution at scale.
Procurement & ecosystem
Added to an existing Microsoft agreement, with identity and compliance integrated out of the box.
Scope and transparency
We would rather set expectations now than during a technical evaluation. This is what Asky supports today.
PostgreSQL, MySQL, Oracle, SQL Server and ClickHouse in production.
Snowflake, BigQuery and Redshift are on the roadmap, prioritized by customer demand.
Questions and answers in any language, with multilingual security guardrails.
Asky generates charts and panels. It is not intended to replace an established enterprise reporting platform.
On-premises via Docker, including air-gapped environments; SSO integrates with your identity provider.
How can we test this safely before committing?
A 30-day evaluation on your own data, measured against criteria you define in advance.
- Connection and scope: three to five tables, one business area
- Documentation preparation: table and column definitions
- Identity and roles: connected to your existing SSO
- Your own questions: a question set defined by your team
- Accuracy measurement: scored with your analysts
- Security validation: access, row and column controls verified
- A measured accuracy result: evidence rather than estimates
- Time-saved analysis: analyst hours per question
- An integration plan, including Copilot surfacing if wanted
Two questions that come up in every buying meeting
What if Microsoft closes this gap itself?
Parts of it will narrow, and we plan for that. What is unlikely to change is the shape of the Microsoft offering: it is designed around Microsoft-hosted data and Microsoft-provided models. Our position sits in the places that assumption does not reach: non-Microsoft operational databases, deployments with no internet route, and customers who must hold their own model keys and encryption keys. If your environment is fully inside the Microsoft cloud, the honest answer is that the gap is smaller for you and we would say so in the first call.
We already pay for Microsoft licences, so how is the extra cost justified?
Asky is priced per business unit and connected data scope rather than per seat, so the cost does not grow with the number of people asking questions. Set it against the work it removes: the modelling and pipeline effort needed to make each new source answerable, the capacity provisioned for it, and the analyst hours currently spent turning questions into SQL. We will size it for your scope on the first call, before an evaluation.
Microsoft product capabilities, licensing tiers and capacity requirements change frequently. Statements about Copilot here reflect our understanding as of July 2026 and should be verified against current Microsoft documentation before any commercial decision.