Miro transformed its collaborative canvas into a hub for human and AI teamwork on May 19, 2026, at the Canvas 26 event in San Francisco. The company announced that its platform now supports shared workspaces where teams, third-party AI agents, internal Sidekicks, and automated Flows work side by side. This move positions Miro as a central nervous system for modern enterprise collaboration, blending visual workflows with artificial intelligence. With over 70 million users worldwide already relying on Miro for brainstorming, planning, and agile ceremonies, the new capabilities aim to dissolve the barriers between human creativity and machine execution.

The Canvas 26 Announcement

Hundreds of product managers, designers, and engineers gathered at Miro’s annual conference to witness the debut of Shared AI Workspaces. CEO Andrey Khusid took the stage to demonstrate how the familiar infinite canvas now accommodates non-human collaborators. “We’re not just adding AI features,” Khusid said. “We’re redefining what a workspace means when agents, automations, and people all contribute to the same surface.” The livestreamed keynote showed a product roadmap evolve in real time as an AI agent suggested dependencies, a Sidekick summarized meeting notes, and a Flow auto-generated Jira tickets from sticky notes.

The announcement reflects a broader industry shift toward compound AI systems that combine large language models with deterministic tools. Miro’s canvas has always been a digital whiteboard; now it doubles as an orchestration layer for distributed intelligence. Early demonstrations revealed integrations with OpenAI’s GPT-5, Microsoft Copilot, and Salesforce Einstein, allowing external agents to read, write, and manipulate board content with proper permissions.

Betas for Shared AI Workspaces roll out immediately to Enterprise plan customers, with general availability expected in Q3 2026. Pricing will be consumption-based, measured in “AI compute minutes” per agent, though Miro confirmed that Sidekicks and basic Flows will be included in existing subscriptions.

Shared AI Workspaces: A New Paradigm

Traditional collaboration tools treat humans as the sole actors. Shared AI Workspaces erase that assumption by giving AI entities their own presence on the board. Each agent or Sidekick appears as a distinct collaborator with an avatar, access rights, and an audit trail. Team members can assign tasks to AI using @mentions, just as they would a human colleague. For example, typing “@AnalyzeBot, summarize this customer feedback cluster” triggers the bot to scan linked stickies and output a structured report directly on the canvas.

Privacy and governance were central concerns addressed at Canvas 26. Administrators can define agent personas, scoping what data each can access and which actions it can perform. Sensitive boards can be restricted to human-only mode, and all AI-generated content is watermarked with its origin. Miro’s Chief Trust Officer emphasized that the company will not use customer data to train third-party models without explicit opt-in—a critical stance given enterprise anxiety around data leakage.

The shared workspace concept extends beyond text-based interactions. Agents can manipulate visual elements: dragging shapes, connecting arrows, and even creating their own frames. In one demo, a design research Sidekick scanned user interview recordings, extracted pain points, and built an affinity diagram autonomously. The team then refined the diagram collaboratively, with the AI updating clusters as new insights emerged.

Third-Party AI Agents Join the Canvas

Miro launched an Agent SDK and marketplace that allow developers to build and list custom AI agents. These agents run in sandboxed environments and communicate with the canvas via a REST API and WebSocket connections. Early partners include Notion AI, which can sync board content with Notion databases; GitHub Copilot Agent, capable of transforming UML diagrams into skeleton code; and Tableau, whose agent inserts live data visualizations into strategic planning boards.

Each third-party agent appears as a standalone collaborator, complete with its own authentication token and activity feed. Teams can subscribe to agents through the marketplace, where usage is metered in AI compute minutes. Miro offers a free tier of 1,000 minutes per month for testing, with additional minutes purchasable at $0.15 per minute. Enterprise contracts include volume discounts and the ability to host agents on-premises or in a private cloud.

Security is multilayered. Agents must be verified by Miro before listing, and enterprise admins maintain a whitelist of approved agents. Data exchanged between the canvas and an agent is encrypted in transit, and agents cannot access boards outside their assigned scope. These provisions aim to satisfy SOC 2, GDPR, and HIPAA compliance requirements, making Shared AI Workspaces viable for regulated industries.

Sidekicks: Miro’s Native AI Assistants

Sidekicks are Miro’s built-in AI assistants, designed for common collaboration patterns. Unlike third-party agents, they run entirely on Miro’s infrastructure and are covered by the platform’s existing data processing agreements. At Canvas 26, Miro introduced three Sidekicks:

  • Summarizer: Condenses long discussion threads, compiles meeting minutes from video call transcripts, and extracts key decisions from sprawling boards.
  • Diagrammer: Converts natural language descriptions into flowcharts, mind maps, or org charts. A product manager can type “Create a user journey from sign-up to first purchase” and watch the Sidekick generate an editable diagram.
  • Facilitator: Acts as a virtual Scrum Master or design sprint moderator, running timed activities, prompting silent brainstorming, and ensuring every participant gets a voice in real-time workshops.

These Sidekicks leverage a combination of proprietary models fine-tuned on collaboration data and frontier models from OpenAI and Anthropic. Miro stated that customer content is never used for training unless the customer opts into the improvement program. Response latency is a critical metric; during the demo, Summarizer processed a 30-page meeting transcript in under 20 seconds, appearing almost instantly on the board.

Sidekicks can invoke Flows—Miro’s new automation engine—to trigger actions beyond the canvas. For example, the Facilitator Sidekick can automatically send Slack reminders to workshop participants or push completed retrospectives to Confluence. This interplay between AI and automation is what Miro calls “intelligent workflow.”

Automated Flows: Streamlining Workflows

Flows replace Miro’s previous automation capabilities with a visual, no-code builder inspired by Zapier and Make. Users define triggers and actions on the canvas itself, connecting Miro to over 200 apps including Jira, Asana, Salesforce, Microsoft Teams, and Figma. A Flow can start when a card moves to a certain feature, when a timer expires, or when an AI agent completes a task.

Canvas 26 demos showcased several high-impact Flows:

  • Agile Workflow: When a user marks a story map item as “ready for development,” the Flow creates a Jira issue, assigns it to the relevant squad, and posts a notification in Teams.
  • Design Handoff: As designers finalize mockups in Figma, a Sidekick extracts the assets and pastes them into a Miro frame, then triggers a Slack message to developers with a link to the design system.
  • Contract Review: A financial services Flow routes a process diagram through Salesforce to gather client data, uses an AI agent to flag compliance risks, and sends a summary PDF to the compliance officer.

Flows run on Miro’s cloud infrastructure, but enterprise customers can choose to execute sensitive steps on their own servers via a self-hosted runner. Execution logs are visible in a new Automation Center, where admins monitor throughput, error rates, and costs. Miro charges for Flows based on execution steps, with a generous free tier of 10,000 steps per month per workspace.

Workflow Automation Meets AI at Scale

The combination of AI agents and automated Flows creates compound automations that adapt to context. Consider a quarterly planning session: The Facilitator Sidekick runs a timeboxed brainstorming exercise; a third-party agent from Tableau pulls live revenue data onto the board; participants vote on priorities using Miro’s polling tool; the Summarizer Sidekick produces a report; and a Flow syncs the final plan with Asana, Jira, and a Confluence page—all without anyone leaving the canvas.

This orchestration is what separates Shared AI Workspaces from simple chatbot integra- tions. Instead of prompt-and-response, the canvas becomes an environment where agents observe changes, react to events, and collaborate over time. Miro’s CTO described it as “moving from stateless AI to stateful, multi-agent systems on a shared substrate.”

Early beta testers reported productivity improvements in cross-functional workflows. A product team at a Fortune 500 company reduced their quarterly planning cycle from three weeks to four days by automating dependency mapping and stakeholder updates. An architecture firm used AI agents to generate code compliance checks from whiteboard sketches, cutting review time by 60 percent.

Integration with Windows and the Microsoft Ecosystem

Miro already enjoys deep integration with Windows, including native desktop apps for Windows 11 with Snap Layouts support and pen input for Surface devices. Canvas 26 brought several Windows-specific enhancements. The Windows Miro app now supports off- line Flows—users can design automations without internet and sync them once reconnected. The app also leverages the Windows Copilot runtime to run Sidekicks locally on devices with NPUs, reducing latency and enhancing privacy.

For Microsoft 365 shops, Miro introduced a bidirectional Microsoft Teams agent. This Sidekick can join Teams meetings, capture whiteboard sketches from the Miro app shared during the meeting, and post summaries directly into the Teams chat. On the reverse, users can @mention the Miro agent in Teams to pull up a board, add a sticky note, or trigger a Flow. The integration uses Microsoft’s AI plugin framework, ensuring compatibility with Purview data governance policies.

OneNote and Loop components also get smarter. A Flow can now embed live Miro boards into OneNote pages, automatically refreshing content as the board updates. In Loop, users can insert Miro components that dynamically reflect the latest AI-generated analyses, creating persistent, real-time dashboards inside collaborative documents.

These integrations position Miro as a bridge between the visual collaboration world and Microsoft’s productivity suite. IT administrators can manage Miro’s AI features through Microsoft Intune, applying conditional access policies and app protection rules. This tight coupling makes Windows the first-class citizen for Shared AI Workspaces, though Miro emphasized that the experience is consistent across macOS, web, and mobile platforms.

Enterprise Implications and Security

For CIOs evaluating Shared AI Workspaces, the governance model will be a decisive factor. Miro’s approach combines centralized administration with granular delegation. Enterprises can create Agent Workspaces that have their own security boundaries, membership, and lifecycle policies. All agent activity is logged in an immutable audit trail searchable from the Miro admin console.

Data residency controls have been extended to AI processing. Organizations can specify that agent inference requests be routed through data centers in specific regions, and Sidekick models can be fine-tuned on a customer’s own data within their Virtual Private Cloud. Miro claims that its architecture isolates agent processes using a microkernel design, preventing cross-tenant data leakage even if an agent is compromised.

Compliance certifications are in progress. Miro expects to achieve FedRAMP Moderate authorization by late 2026, opening the door for U.S. government agencies to adopt AI-powered whiteboarding. Existing certifications covering ISO 27001, SOC 2 Type II, and HIPAA remain in effect for the core platform, with AI-specific audits under way.

Analysts covering the announcement noted that while Miro’s vision is compelling, execution risk remains high. Managing stateful, multi-agent systems at scale introduces challenges around conflict resolution, cost predictability, and model drift. Miro has published a roadmap addressing these concerns, including a Conflict Resolution Sidekick that will reconcile contradictory agent outputs and a Cost Explorer dashboard projected for Q4 2026.

The Road Ahead for Collaborative AI

Canvas 26 marks Miro’s most ambitious platform shift since its founding in 2011. By making the canvas a shared space for diverse intelligences, the company is betting that visual collaboration tools can become the operating system for modern teamwork. If successful, Shared AI Workspaces could obsolete the patchwork of chat, documents, and project management tools that fragment today’s workflows.

Competitors are not standing still. Figma’s AI features continue to evolve, and Microsoft Whiteboard is integrating deeper Copilot capabilities. However, Miro’s open agent architecture and robust automation engine give it a head start in multi-agent orchestration. The true test will come when thousands of agents operate simultaneously across global enterprise deployments—a scenario Miro’s infrastructure team is actively stress-testing.

For Windows users, the immediate takeaway is a more intelligent collaboration surface that works seamlessly with the tools they already use. The ability to run Sidekicks locally on NPU-powered Windows devices offers a taste of hybrid AI architectures that balance cloud scale with edge privacy. As AI copilots become ubiquitous, Miro’s canvas provides a shared context where those copilots can truly collaborate. The era of solo AI is giving way to team AI—and the whiteboard is its home.