01
Microsoft redesigns Copilot around Home, Code and Autopilot
Microsoft announced Home for Chat and Cowork, Code for natural-language solution building, and Autopilot for persistent, proactive task execution. Home and Code will roll out through Frontier, while Microsoft plans to expand Autopilot’s private preview.
- Keep initial deployment plans limited to pilots until Microsoft confirms availability and licensing for your tenant.
- Define delegated permissions, data-access boundaries and human-approval requirements before testing unattended Autopilot tasks.
- Assess Copilot Managed Runtime’s hosting and security controls before allowing employees to deploy AI-built applications or agents.
02
Fabric IQ brings governed business definitions into Copilot
Microsoft made Fabric IQ generally available in Copilot Chat and Cowork. The integration grounds responses in governed Power BI semantic models, metrics, relationships and business definitions.
- Prioritize semantic models with named business owners and agreed financial or operational definitions for initial Copilot use cases.
- Test whether existing model permissions prevent employees from retrieving restricted business information through Copilot.
- Validate Czech-language questions, data-location requirements and answer accuracy before extending access beyond a pilot group.
03
Azure Local adds production SQL and preview AI inference
Microsoft made SQL Server on Azure Local generally available for connected and disconnected operations. It also highlighted Foundry Local on Azure Local in preview for running AI inference near SQL Server data.
- Separate the generally available database deployment from the preview inference capability in procurement and production-readiness decisions.
- Evaluate the local architecture for manufacturing or restricted-connectivity sites without assuming it automatically satisfies regulatory obligations.
- Request supported hardware, model availability, licensing eligibility and operating-cost details before comparing it with Azure-hosted inference.
04
Copilot gets a centralized plugin registry
Microsoft announced a registry for discovering, publishing and governing plugins across Copilot experiences, Microsoft 365 apps and SharePoint. The registry includes more than 100 plugins containing skills, connectors and agents.
- Assign ownership of plugin approval across M365 administration, security and procurement before opening discovery to employees.
- Review publisher trust, connector permissions and access to corporate data for each proposed plugin.
- Test user- and group-level availability controls against your existing departmental access policies.
05
Copilot adds inline agent and skill invocation
Microsoft’s September Copilot update introduced direct invocation of agents with “/” and skills with “@” inside prompts.
- Update internal Copilot training to distinguish agent invocation from skill invocation.
- Use clear names and descriptions so employees can identify approved departmental capabilities.
- Test invocation with restricted user accounts to check that access policies still apply within combined workflows.
06
Office and Teams move under Microsoft’s Copilot organization
Microsoft announced that Office and Teams teams will move into its Copilot, Agents, and Platform organization.
- Bring Office, Teams and Copilot owners into the same internal roadmap review to coordinate adoption and change management.
- Ask Microsoft account teams for product-specific commitments rather than treating the reorganization as a release roadmap.
- Keep licensing and migration assumptions unchanged until Microsoft publishes concrete product or contractual changes.
07
VS Code expands Copilot’s agent-driven development workflows
GitHub’s September VS Code release summary added scheduled automations, agent merge, pull-request creation from agent sessions and richer GitHub context. It also introduced HydraFusion as a research preview for automatic model and workflow selection.
- Pilot agent-created pull requests with existing branch protection and mandatory human review intact.
- Check repository permissions and execution credentials before enabling scheduled engineering tasks.
- Evaluate HydraFusion separately from established workflows because its research-preview status requires additional validation.
08
GitHub previews code-defined, multi-agent Copilot workflows
GitHub introduced dynamic workflows in public preview across Copilot CLI, the Copilot app and the Copilot SDK. Developers can define sequential and parallel orchestration in code.
- Start with a repeatable internal development task whose inputs, outputs and acceptance criteria are explicit.
- Review credentials and data access across every workflow step before connecting internal systems.
- Test failure handling, auditability and support limitations before considering production deployment.
09
GPT-6.1 Sol becomes generally available in GitHub Copilot
GitHub made OpenAI’s GPT-6.1 Sol generally available in Copilot, with rollout across Visual Studio Code, Visual Studio, CLI and coding-agent experiences.
- Use Business and Enterprise model policies to stage access rather than enabling the model across all development teams immediately.
- Compare code quality, token consumption and task completion against your approved models using representative internal repositories.
- Check availability in each team’s IDE and agent workflow before standardizing training or tooling around the model.
10
Claude Sonnet 5.5 reaches general availability in GitHub Copilot
GitHub made Claude Sonnet 5.5 generally available for eligible Copilot plans, including Business and Enterprise. The gradual rollout covers Visual Studio Code, Visual Studio, CLI and coding-agent experiences.
- Confirm tenant availability before scheduling team-wide evaluations because the rollout is gradual.
- Control access centrally through Copilot model policies and include the model in your approved-model inventory.
- Benchmark Czech-language requirements, generated tests and repository-specific coding conventions alongside latency and cost.
11
Copilot code review gains APIs and a new default
GitHub added REST and GraphQL support for requesting Copilot code reviews and selecting review effort. Balanced became the default for new and existing repositories and organizations unless they had explicitly selected Lite.
- Audit organization and repository settings to identify where the new default changes existing review behavior.
- Use the APIs to test AI review within established CI/CD approval gates rather than replacing human sign-off.
- Measure review duration, useful findings and resource consumption before selecting an effort level for wider deployment.
12
Microsoft reports global generative-AI adoption reached 18.8%
Microsoft’s Q2 2026 Global AI Diffusion Report estimated that 18.8% of the global working-age population used generative AI in June, up from 17.8% in the first quarter. The report also identified a widening adoption gap between the Global North and Global South.
- Do not use the global working-age adoption rate as a direct benchmark for paid Copilot usage inside a Czech enterprise.
- Measure internal active use by role and department before allocating additional licenses or training budgets.
- Review the report’s methodology and geographic coverage before using its figures in board-level adoption targets.