What is new in copilot studio and you need to know everything

Modern copilot studio with new canvas workflow design

​Copilot Studio has recently undergone a massive shift toward "agentic" capabilities, moving beyond simple chatbots to autonomous agents that can reason and act.

​As of early 2026, here are the most significant updates for agentic flows:

​1. New "Workflows" (Public Preview)

​Microsoft recently introduced a redesigned visual canvas specifically for Workflows. This is a dedicated, agentic-led way to build automation.

• ​Deterministic Execution: Unlike open-ended chat, these follow a rule-based path where the same input always produces the same output.

• ​Native AI Actions: You can now drag and drop native AI capabilities—like summarization or data extraction—directly into the flow.

• ​Agent Handoffs: A workflow can now pause and hand off a task to another specialized agent or a human, then resume once the task is complete.

​2. "Computer Use" Agents (CUA)

​One of the most advanced "agentic" features is the ability for agents to interact with legacy software that lacks an API.

• ​UI Navigation: Agents can "see" a screen, click buttons, and type into fields on a hosted Windows 365 Cloud PC.

• ​Use Case: This is ideal for tasks like entering data into old ERP systems or scraping market research from websites that don't have structured data feeds.

​3. Multi-Agent Orchestration (A2A)

​Copilot Studio now supports Agent-to-Agent (A2A) communication.

• ​Specialization: Instead of one giant, "know-it-all" agent, you can build several "sub-agents" (e.g., one for pricing, one for technical docs).
• ​Delegation: A "Primary Agent" can now intelligently delegate parts of a complex user request to these specialized sub-agents and combine their answers into a single response.

​4. Advanced "Agentic RAG"

​Retrieval Augmented Generation (RAG) has become more intelligent:

• ​Contextual Reading: Agents no longer just "search" for keywords; they can intelligently decide whether to read an entire SharePoint file or just specific sections based on the query.

• ​Comparison: They can now compare multiple files (like two different versions of a policy) directly within SharePoint to highlight differences.

​5. Human-in-the-Loop (HITL) Integration

​To keep agentic flows safe, Microsoft added the "Request Information" action.

• ​Approval Breaks: If an agent reaches a high-stakes decision (like approving a $10,000 refund), it can automatically send a structured request to a human via Outlook.

• ​Dynamic Resumption: The flow pauses until the human responds, and then uses that human input as a parameter to continue the process.

​Pro Tip: To access many of these "agentic" features, ensure you are using the preview environment via preview.microsoft.com, as many of the advanced visual designers are currently in public preview.

Knowledge Agent in SharePoint: What It Is, How It Works, and How to Set It Up

What is the Knowledge Agent in SharePoint?

The term Knowledge Agent in SharePoint generally refers to an AI-powered assistant that uses your SharePoint content to answer questions, surface insights, and streamline knowledge discovery while respecting permissions. In practice, this is often implemented with Microsoft 365 Copilot, Microsoft Search, and optional add-ons like Viva Topics and SharePoint Premium to organize, retrieve, and generate responses grounded in your SharePoint sites, libraries, and lists.

Why organizations use a Knowledge Agent in SharePoint

• Faster answers: Teams get instant, permission-trimmed answers from policies, SOPs, and project docs.
• Reduced duplicate work: Surfaces existing assets so people reuse content instead of recreating it.
• Consistent knowledge: Standardizes responses based on authoritative sources and metadata.
• Better onboarding: New hires find tribal knowledge and how-to guidance quickly.

How a Knowledge Agent in SharePoint works

• Grounded retrieval: Uses Microsoft Search and Graph signals to find the most relevant SharePoint items the user can access.
• Security trimming: Answers are constrained by the user’s existing permissions; blocked content is never exposed.
• Metadata and taxonomy: Columns, content types, and terms improve ranking, relevance, and summarization quality.
• Optional enrichment: Viva Topics builds topic pages; SharePoint Premium (formerly Syntex) can auto-classify and extract metadata.

Common scenarios and example prompts

Policy and compliance

Ask: “Summarize our travel reimbursement policy and list required receipts.” The agent retrieves the latest policy page or PDF from the HR site and provides a concise, cited summary.

Project knowledge

Ask: “What are the milestones and risks for Project Orion?” The agent compiles milestones from a SharePoint list and risks from a project wiki, linking back to the sources.

Customer support

Ask: “How do I troubleshoot a failed connector?” The agent surfaces a step-by-step SOP from a knowledge library and highlights escalation paths.

Setting up a Knowledge Agent using SharePoint as the knowledge base

• Confirm data foundations: Store authoritative documents in SharePoint with clear naming, versioning, and owners.
• Structure content: Use content types, columns, and taxonomy for policies, procedures, and FAQs.
• Enable enterprise search: Ensure SharePoint content is indexed and accessible via Microsoft Search.
• Optional Copilot configuration: If you use Microsoft 365 Copilot or Copilot Studio, connect SharePoint sites as data sources so the agent can retrieve and ground answers.
• Define scope and guardrails: Limit the agent to curated sites and libraries; maintain a whitelist of trusted sources.
• Pilot with a team: Start with HR, Finance, or Support to test quality, then expand organization-wide.

Best practices for high-quality answers

• Keep content current: Archive superseded documents and set review cadences (e.g., quarterly).
• Standardize titles and summaries: Add executive summaries and clear titles for better retrieval and summarization.
• Use templates: Consistent templates for SOPs, FAQs, and runbooks improve answer reliability.
• Govern metadata: Apply required columns (owner, effective date, version) and managed terms.
• Citations and links: Ensure the agent returns links to source files so users can verify details.
• Measure and iterate: Track unanswered queries and refine content to close gaps.

Security, compliance, and governance

• Respect permissions: The agent inherits SharePoint and Microsoft 365 permissions; avoid broad site access unless necessary.
• Label sensitive content: Use sensitivity labels and DLP policies to prevent oversharing.
• Audit and monitoring: Review logs and analytics to ensure the agent performs as intended.

Troubleshooting relevance and quality

• Low-quality answers: Improve source documents, add summaries, and use clearer titles/headers.
• Missing files: Confirm search indexing is enabled and the site/library is in scope.
• Outdated information: Retire old versions and highlight the latest approved document.
• No citations: Prefer storing authoritative content in SharePoint pages or modern libraries with metadata and avoid scattered personal file shares.

Frequently asked questions

Does the Knowledge Agent access everything in SharePoint?

No. It only accesses what a user is already permitted to see, honoring security trimming.

Do we need Viva Topics or SharePoint Premium?

Not required, but they enhance organization and metadata extraction, which can improve answer quality.

Can we limit the agent to specific sites?

Yes. Scope the agent to selected SharePoint sites and libraries to keep answers trustworthy and on-topic.

How do we keep knowledge fresh?

Assign content owners, add review schedules, and monitor unanswered queries to guide updates.

Getting started

Identify your top knowledge scenarios, curate authoritative SharePoint libraries, and pilot a scoped Knowledge Agent in SharePoint. With strong information architecture and governance, you’ll deliver faster, more accurate answers at scale—without compromising security.