Google is reportedly preparing to test “Project Jarvis,” an intelligent assistant integrated within the Chrome browser, according to The Information. Set for a December release to select users, Jarvis is said to run on Google’s latest Gemini 2.0 model, assisting with online tasks such as organizing information or booking reservations within the Google Chrome browser. The AI agent‘s primary function would be to handle repetitive tasks directly on the web, showcasing Google’s approach to making digital routines simpler for users within its browser ecosystem.
Anthropic’s Claude 3.5 Sonnet Adds Desktop Automation
Anthropic, one of the other leading AI players, recently launched an update to its Claude 3.5 model, introducing a “Computer Use” feature that expands AI capabilities to interact directly with a user’s desktop. Claude can perform commands like typing, clicking, and handling multiple applications on PCs, which opens up possibilities for companies looking to automate data entry, document organization, or customer service routines. This feature is currently available through Anthropic’s APIs or on Amazon Bedrock and Google Cloud’s Vertex AI, enabling flexible integration for businesses.
By enabling Claude to execute tasks traditionally requiring human interaction, Anthropic provides a distinct alternative to web-only agents like the rumored Google’s Jarvis, positioning itself in desktop environments with added versatility. The Sonnet update’s “Computer Use” mode captures and interprets screen elements similarly to Google’s Jarvis but is less restricted by browser limitations, allowing more robust desktop-level functionality.
How Jarvis Works in Chrome and Why It’s Different
Unlike Anthropic’s desktop-oriented Claude Sonnet, Google’s Jarvis is said to operate within Chrome, acting as a digital assistant to navigate online workflows. Jarvis analyzes screenshots of web pages to interpret commands, then follows up with clicks or data entry actions based on those interpretations. Although in early development, Jarvis’s current functionality hints at a future where it could take over mundane tasks that users currently do themselves. Google’s move to embed Jarvis within Chrome aligns with its goal of enhancing user experience directly in the browser, eliminating the need for external software installations.
While Anthropic’s Claude manages tasks across desktop applications, Google’s browser-based approach would prioritize web-specific functions, making it a potential asset for users seeking automated assistance in an already familiar setting. Gemini 2.0, the AI engine behind Jarvis, would bring Google’s advanced language processing capabilities into this tool, providing contextual task handling and navigation within the Chrome browser.
Microsoft Joins the Agent Race with Dynamics 365 Copilot Agents
Meanwhile, Microsoft is advancing its AI agent capabilities within enterprise applications, specifically through new agents for Dynamics 365 unveiled also in October. Designed for industries from finance to sales, these agents automate routine tasks within specialized domains. For instance, the Sales Qualification Agent categorizes leads and customizes outreach, while the Financial Reconciliation Agent cleans up data inconsistencies in transaction records. Unlike Jarvis or Claude, Microsoft’s agents are highly specialized for business applications, aimed at improving productivity within companies rather than individual users.
Microsoft’s Copilot Agents, initially introduced at the Build 2024 conference, work across platforms like Teams and Outlook, embedding automation within established business software. Copilot Studio, an extension of this feature, lets organizations design custom workflows tailored to their unique requirements, adding a layer of flexibility beyond predefined tools. Early adopters like Vodafone and Cognizant have integrated these AI agents at scale, reporting noticeable improvements in productivity.
Microsoft is also expanding general Copilot features across its own services, including integrations with tools like SharePoint, OneDrive, and recently even Bing Webmaster Tools. These agents now assist with analysis, creating summaries, meeting organization, document management, and project coordination, transforming Copilot from a reactive tool into an AI partner in day-to-day operations. This effort echoes similar initiatives from Google and Anthropic, each working to establish autonomous AI tools across different contexts.
While Salesforce CEO Marc Benioff recently heavily criticized Microsoft’s Copilot as “Clippy 2.0”, basically saying it was useless, large enterprises like already have moved forward with significant Copilot investments, securing tens of thousands of licenses to streamline various workflows. Copilot’s customizable interface, accessible through Copilot Studio, allows companies to refine agents to meet precise departmental needs, marking a clear contrast from more consumer-focused assistants like it would be Google Jarvis.
So, Google, Anthropic and Microsoft are shaping the AI agent landscape with complementary, but distinct, approaches. While Anthropic’s Claude series prioritizes flexibility across devices, Google’s Jarvis seems to align with its web-centric ecosystem, and Microsoft’s Copilot agents remain firmly rooted in business automation. Each company’s strategy reflects different use cases.