Database administrators who depend on AI assistance inside SQL Server Management Studio will find the first public preview of SSMS 22 conspicuously silent: Microsoft has confirmed it will ship without any Copilot functionality. The removal marks a sharp strategic pivot—the company is halting its in‑house Copilot that ran on Azure OpenAI and will instead integrate GitHub Copilot down the road, a move triggered by overwhelming user feedback that rejected a separate cloud‑provisioning nightmare.

The Short‑Lived SSMS 21 Copilot Experiment

When Microsoft shipped Copilot in SSMS as an optional extension for SSMS 21, it promised to help developers write, debug, and explain Transact‑SQL through natural language. The assistant could tap into the active database connection—reading the SQL Server version, schema objects, and indexes—to generate context‑aware responses. Behind the scenes, it used retrieval‑augmented generation (RAG) to blend model outputs with database metadata.

But the setup was anything but simple. To use the preview, administrators had to provision an Azure OpenAI endpoint, deploy a supported model (Microsoft’s docs recommended gpt‑4o), configure authentication via Entra roles or API keys, and manage rate limits and content filters. That multi‑step process became a major friction point. “We expected that some users would encounter friction during initial configuration,” acknowledged Erin Stellato, principal program manager at Microsoft, in a Tech Community post.

Despite the complexity, the feature had genuine enterprise chops. The preview documentation made clear that Copilot did not retain prompts or use customer data to train models, and it operated under the user’s permissions when running queries—critical for compliance. Still, many database professionals balked at having to link a core administrative tool to a separate cloud AI spend.

User Feedback Forced a Rethinking

Less than four months after public availability, the SSMS team ran a survey to gauge user sentiment. According to press reports, 75 percent of respondents wanted the Copilot experience in SSMS to leverage GitHub Copilot—not a standalone Azure OpenAI deployment. While the exact survey breakdown hasn’t been published by Microsoft, the directional signal was strong enough to alter the product roadmap.

“At the end of the day, the data – combined with user feedback – supported the move to using GitHub Copilot,” Stellato wrote. “And that’s what we’re doing.” The SSMS team sees this as creating a unified AI experience across the Microsoft ecosystem, aligning with how developers already use GitHub Copilot in Visual Studio and VS Code.

What’s Missing in SSMS 22 Preview 1—and Why It Matters

With the pivot, SSMS 22 Preview 1 will have no Copilot at all. There’ll be no option to install AI components, no Copilot menu entry, and no AI-generated suggestions. For DBAs who had grown accustomed to the SSMS 21 assistant, this is a stark regression.

That removal is more than cosmetic. The original Copilot’s value came from its deep database context: it knew the server type, the active schema, and could tailor T‑SQL suggestions accordingly. GitHub Copilot, designed as a general‑purpose coding assistant, currently lacks that contextual awareness. “The problem is that some of the most significant values of Copilot in SSMS include its connection, database context, and its retrieval-augmented generation implementation,” Stellato said. “The current GitHub Copilot integration needs these to be added.”

Microsoft concedes that feature parity will take multiple releases. Until then, a DBA asking GitHub Copilot to generate a query against a specific table might get syntactically valid SQL that ignores real‑world constraints or uses deprecated syntax. The assistant won’t see indexes, statistics, or server‑level properties—a huge gap for performance‑sensitive work.

The GitHub Copilot Integration: Promises and Pitfalls

Aligning with GitHub Copilot offers clear benefits. Developers already subscribe to it, so they expect the same license to work inside SSMS without provisioning extra cloud resources. The unified experience also simplifies training and governance, since the assistant’s behavior, keyboard shortcuts, and trust boundaries would be consistent across tools.

But several technical hurdles stand in the way. To match the older preview, the GitHub Copilot integration must:

  • Gain secure access to the active database connection’s schema, server version, and object metadata.
  • Implement a RAG pipeline that pulls local database context into prompts without leaking sensitive metadata.
  • Enforce the logged‑in user’s permissions when executing or suggesting queries, with strict read‑only defaults.
  • Provide audit logs that show prompts, suggested SQL, and execution approvals—vital for regulated industries.

Microsoft has not yet detailed how it will address these points. The original Azure OpenAI integration gave administrators explicit control over endpoints, keys, and content filters; any replacement must match or exceed that transparency, especially since GitHub Copilot’s backend and model‑hosting choices may differ from Azure OpenAI.

Licensing and Procurement Headaches Loom

The shift also raises unanswered commercial questions. Will existing GitHub Copilot seats cover SSMS usage without additional cost? Does the GitHub Copilot Business or Enterprise license model satisfy corporate procurement policies, or will organizations need separate purchases? The SSMS team has acknowledged user concern but hasn’t provided clarity.

This ambiguity matters because procurement cycles are slow. If the final answer requires new licensing, large enterprises could face months of negotiations before DBA teams can legally use the AI features. Some may be tempted to stick with SSMS 21 as a “feature‑frozen preview” indefinitely.

Security and Governance: A Higher Bar for Database Tools

Database environments introduce security demands that go beyond code editors. An assistant that can see schema information or propose executable queries must undergo rigorous risk assessment. Even with the SSMS 21 preview’s explicit controls—role‑based access to the Azure OpenAI endpoint, non‑retention of prompts for model training—organizations had to create internal policies around what queries could be auto‑generated and who could enable read/write modes.

Any GitHub Copilot integration must meet the same bar. Enterprises will expect Microsoft to publish detailed transparency notes confirming that prompts are not used for model training, that data stays within contractual boundaries, and that strong authentication and authorization mechanisms are in place. The absence of such guarantees could block adoption in regulated sectors like finance and healthcare.

What Database Teams Should Do Now

For IT leaders managing SSMS deployments, the immediate priority is preparation:

  • Inventory current Copilot usage. Identify who relies on SSMS 21’s AI features and for which tasks—NL‑to‑SQL, query fixing, schema exploration.
  • Decide on a migration stance. If those features are mission‑critical, staying on SSMS 21 (feature‑frozen but still available) is a viable stopgap. Teams that can wait should plan to skip AI assistance in upcoming SSMS 22 previews and monitor the roadmap.
  • Document any Azure OpenAI configurations. For those who tested the old preview, keep records of endpoints, deployments, token limits, and role assignments; they’ll be useful when evaluating the future GitHub Copilot integration.
  • Establish an AI assistant governance policy now. Define which query types can be auto‑generated, mandatory human review for schema‑changing operations, and how prompts and results will be logged and audited.
  • Test any assistant in staging first. No AI‑generated SQL should touch production without thorough validation against representative schemas and permissions.
  • Subscribe to official SSMS channels. Microsoft’s Tech Community blog and the SSMS release notes are the best sources for announcements on GitHub Copilot integration timelines and licensing details.

Conclusion

Microsoft’s decision to remove Copilot from SSMS 22 Preview 1 and replatform around GitHub Copilot is a pragmatic response to user frustration, but it leaves database professionals in a temporary void. The short‑term trade‑off is clear: fewer setup headaches at the cost of losing DB‑specific AI smarts that made the original preview valuable. Whether the eventual GitHub Copilot integration can match—and exceed—those capabilities depends on Microsoft delivering deep schema awareness, robust security guarantees, and a licensing model that aligns with enterprise expectations.

For now, DBAs must choose between holding onto a frozen preview or adopting a tool without AI assistance. The architecture and governance decisions made during this transition will determine whether Copilot becomes a trusted productivity booster for SQL Server teams or just another source of compliance friction.