Microsoft is testing the waters for a radical overhaul of Edge's Copilot, one that could transform the browser from a simple summarizer into an autonomous agent capable of completing tasks across multiple tabs and applications. A new user survey, first spotted by Windows Latest, asks Edge users about advanced capabilities like multi-tab context, extracting data into Excel, drafting and sending emails, and filling forms automatically—features that would bring Edge closer to Perplexity's Comet, an AI-first browser built around agentic workflows.
The survey's questions are not mere curiosity; they signal a strategic pivot. Microsoft appears ready to move beyond the current Copilot mode, which primarily answers questions and summarizes page content, toward a future where the assistant can act on the user's behalf. This would mark a significant leap for Edge, embedding deep automation into daily browsing while leveraging the browser's unique integration with Windows and Microsoft 365.
What Edge Copilot Can Do Today—and What It Might Do Tomorrow
Copilot mode in Microsoft Edge, introduced in preview earlier this year, replaces the traditional New Tab Page search box with a Copilot-centric compose box and adds an "Ask Copilot" integration in the address bar. It can read the active tab's content to provide summaries or answer contextual queries. While handy, it stops short of true task automation. You can ask Copilot to compare products on a page, but it won't switch tabs, pull data into a spreadsheet, or compose an email in Outlook.
The survey suggests Microsoft wants to change that. It probes user appetite for multi-tab context—the ability for Copilot to reason across all open tabs, not just the current one. This would allow a single prompt to aggregate information from multiple sources, like compiling a unified shopping list from several retailer sites or comparing specs across product pages. More ambitiously, the survey explores direct actions: extracting structured data (such as tables) into Excel, drafting an email from page content and sending it, auto-filling forms using aggregated tab context, and generating consolidated lists.
These capabilities mirror the agentic promise of Perplexity Comet, a browser built from the ground up around autonomous agents and persistent session memory. Comet's design treats AI not as an add-on but as the operating logic, enabling it to carry out multi-step tasks with minimal user intervention. By surveying users, Microsoft is gauging whether Edge's massive installed base is ready to embrace similar autonomy—and whether it can be monetized.
Monetization Signals: Copilot Pro and the Hidden "Journeys" Feature
The survey doesn't just explore features; it tests pricing tolerance. Windows Latest reports that some upcoming advanced AI capabilities in Edge could be locked behind a Copilot Pro subscription, priced around $20 per month. Among these is a feature called "Journeys," which uses local AI models—not cloud-based Copilot—to summarize and group browsing activity. Journeys promises to organize a user's web history into coherent narratives, helping pick up research threads without manual bookmarking. The survey asks about usage frequency and specific verticals like travel, shopping, and coding, suggesting Microsoft may tune prompts and UI flows for each domain, a common pre-launch step before paid gating.
Tying Copilot Pro to productivity tasks like Excel extraction or email drafting aligns with Microsoft's broader strategy to monetize AI across its ecosystem. For users deeply embedded in Microsoft 365, the value proposition is clear: the ability to move from research to action without leaving the browser. However, paywalling core automation risks alienating mainstream users who may view such features as fundamental to a modern browser.
Technical Realities: How Microsoft Might Implement Multi-Tab Agentic Abilities
Building a browser that can "see" and act across dozens of tabs is no small feat. Large language models have finite context windows; feeding them raw content from 20 tabs would quickly hit limits and degrade performance. Microsoft will likely rely on hybrid architectures: local models for initial summarization and context compression, then cloud LLMs for complex reasoning. This approach is already hinted at in experimental Edge features, such as local AI for browsing history summarization in Journeys.
Local processing offers additional benefits: reduced latency and stronger privacy guarantees. A sensitive task like filling a medical form could be handled partially on-device before any cloud interaction, leaving users in control of what gets sent to Microsoft's servers. But for agentic features that require credential use—like sending emails or booking reservations—secure handling becomes paramount. Microsoft's existing approach, which requires explicit user opt-in and visual indicators when Copilot accesses content, will need extension. Expect tokenized access, scoped permissions, and multi-factor confirmations for sensitive actions.
The Risks: Privacy, Errors, and Subscriptions
Any feature that scans multiple tabs or interacts with personal accounts raises red flags. Users and enterprises will demand granular controls: per-task permissions, clear disclosure of data flows, and an option to run entirely local when desired. Without these, trust erosion is likely, especially among business and privacy-conscious users. The survey's emphasis on permissions suggests Microsoft is aware, but the proof will be in the final UX.
Automation errors pose another danger. If Copilot can draft and send emails, book flights, or fill payment forms, a mistake could lead to embarrassment, financial loss, or legal liability. Microsoft will need conservative defaults, undo capabilities, and transparent activity logs. Security researchers will also scrutinize prompt injection risks: could a malicious page trick Copilot into performing unintended actions? Robust input sanitization and adversarial testing are non-negotiable.
Then there's the subscription model. A $20 Copilot Pro tier might be palatable for power users and enterprises, but it could fragment Edge's feature set. When Apple Intelligence or Google's AI features are increasingly free, Microsoft risks being seen as nickel-and-diming its own users. The challenge is delivering enough value to justify the cost without making the free tier feel anemic.
A Blueprint for Responsible Agentic Browsing
If Microsoft gets this right, the following safeguards should be non-negotiable:
- Scoped permissions: Users should grant access per task type (e.g., "allow Copilot to read open tabs for summaries") with a dashboard for revocation and audit.
- Progressive action confirmation: Read-only context by default; explicit single-use consent for actions like sending emails; multi-factor steps for financial tasks.
- Local-first architecture: An on-device summarizer to compress content before any cloud call, plus a "local only" mode for maximum privacy.
- Explainability and undo: Every automated action should come with a short rationale and a clear reversal path.
- Enterprise controls: Group Policy and MDM settings to globally enable or disable agentic features, plus audit logs for compliance teams.
These measures would go a long way toward addressing legitimate concerns while letting users benefit from automation.
Why Microsoft Is Making This Move Now
The competitive pressure is undeniable. Google, OpenAI, and Perplexity are redefining search and browsing with AI-first interfaces. By reshaping Edge as an agentic platform, Microsoft can re-center users in its ecosystem and create sticky integrations with Office, Teams, and OneDrive. A browser that automates data transfer between web pages and Excel, for instance, becomes a productivity multiplier that competitors without that office suite tie-in will struggle to match.
The survey-driven approach also indicates a desire to calibrate features based on real user feedback. Microsoft isn't simply copying Comet; it's testing which verticals (travel, shopping, coding) offer the highest impact and monetization potential. This data will likely inform not only feature design but also prompt tuning and UI flows tailored to each domain.
Practical Guidance for Users and IT Administrators
For regular users, treat upcoming Copilot automation as a productivity accelerator—but keep confirmations on for any action that posts or sends data. Use local-only or opt-out settings when handling sensitive information.
Power users should evaluate ROI if advanced features land behind Copilot Pro. Test workflows with non-critical data first and verify the audit trail. For IT administrators, demand granular MDM controls and auditability before enabling agentic features in corporate environments. Create policies that require tokenized credential access to minimize risk.
What We Don't Know Yet
Despite the survey's hints, several details remain unconfirmed. Exact pricing and feature gating for Copilot Pro are provisional; Microsoft has not announced final packaging. The full scope of automation—how proactive Copilot will be versus how much it will still require user prompts—is unclear. And the technical stack, particularly the split between local and cloud processing, is still speculative based on exploratory reports. As with any pre-release intelligence, these plans could change.
The Road Ahead: A Phased Rollout
Microsoft is already testing Copilot-mode improvements in Canary and controlled rollouts. Expect a phased progression: broader availability of improved summarization and multi-tab prompts, experimental "Actions" behind feature flags, and eventually premium gating for the most agentic features. Enterprise controls and local-model options will likely follow as security and regulatory concerns are addressed.
The survey is more than market research. It's a roadmap hint that Edge could evolve from a context-aware assistant into a true agentic browser capable of multi-tab reasoning and cross-application automation. That vision brings Edge closer to Perplexity Comet's functional model while leveraging Microsoft's unique strengths: deep Windows integration and Office interoperability. The opportunities are compelling—faster research, one-click synthesis of multi-tab projects, seamless transfer of web data into productivity apps. But the risks are real: privacy, security, and the user-experience cost of gating useful features behind a subscription.
For users and administrators, cautious optimism is the best posture. Explore Copilot's new capabilities where they help, demand transparent consent and auditability where they act, and insist on fine-grained controls at the enterprise level. As Edge's Copilot-first experiment matures, the balance Microsoft strikes between automation power, user control, and trust will determine whether agentic browsing is seen as a leap forward or a convenience many will accept only with careful safeguards.