Google is integrating Gemini-powered automation into Chrome to enhance user workflows, though experts suggest it’s best suited for tasks with minimal risk.
Google aims to broaden Chrome’s utility in corporate environments with a new auto-browse feature, powered by its Gemini 3 model. This innovation is designed to navigate websites, collect data, and process information, thereby reducing the need for manual data entry and repetitive clicking in professional tasks.
The feature is currently available for preview to paying AI Pro and Ultra subscribers in the US, accessible via the Gemini interface within Chrome.
This initiative comes as major cloud providers and AI model developers like OpenAI and Anthropic increasingly embed artificial intelligence into business operations. Their goal is to automate routine tasks and processes, leading to significant productivity improvements.
OpenAI led the charge in this area in February 2024, demonstrating software capable of autonomous device operation. Anthropic followed suit in October of the same year, showcasing its “computer use” functionality. Since then, both firms have refined these capabilities into more sophisticated products, offered either as standalone solutions or, in Anthropic’s case, integrated directly into its generative AI chatbot, Claude.
Google has also been exploring browser and agent-based automation for some time, including Jarvis, first revealed in October 2024 and now known as Project Mariner. This project investigates more independent web navigation and task execution, aligning with Google’s wider strategy to establish AI as an active contributor to enterprise operations.
Jarvis was unveiled merely days after Anthropic’s computer use demonstration and is currently accessible as a prototype to subscribers of the $250/month Google AI Ultra service.
Currently in Preview
Chrome’s auto-browse function is accessible as a preview in the US for AI Ultra subscribers, and also for those subscribed to the $20/month AI Pro service. Analysts view this as a strategic move to position the browser as a versatile productivity tool, designed to simplify knowledge-based work and relieve employees from mundane, repetitive online activities.
Avasant principal analyst Abhisekh Satapathy commended Google’s integration of user supervision, noting that Gemini prompts for confirmation before executing specific actions.
Pareekh Jain, principal analyst at Pareekh Consulting, highlighted its user-friendliness when dealing with intricate workflows.
“It possesses the capability to manage intricate multi-step web processes, such as form completion and site navigation. Its enterprise applications span expense report processing (by extracting receipts from portals), aggregating procurement quotes from various vendor websites, and updating CRM systems via SaaS interfaces,” he explained.
Development teams could experience significant productivity enhancements, he added: “This innovation could unlock substantial benefits through zero-code automation, empowering operational teams in HR or Finance to independently set up small automations – for instance, instructing it to visit the vendor portal, download January invoices, and save them to a specified Drive folder, all without developer intervention.”
Jain remarked that by being relieved of such monotonous tasks, “Developers can then transition from creating brittle web scraper scripts to authoring high-level agentic instructions, shifting their focus from individual clicks to achieving desired outcomes, thereby boosting efficiency across various workflows.”
Everest Group practice director Priya Bhalla commented that Chrome auto browse might fundamentally alter how developers approach user experience design: “In the long run, this could change how developers conceive UX – optimizing not only for human users but also for AI agents operating on their behalf.”
Potential Risks Identified
Nonetheless, analysts cautioned that Chrome’s auto-browse feature might not be suitable for critical workflows.
Enterprise systems frequently incorporate layers of authentication, role-based access controls, conditional logic, and customized interfaces, which Jain believes are areas where Chrome’s auto-browse could encounter difficulties.
“It relies exclusively on browser interactions without deep API or internal system integration. Additionally, it could prove unstable on dynamic web pages susceptible to DOM changes,” Jain elaborated.
Typically, an agent utilizes the Document Object Model (DOM) to navigate a webpage, as the DOM outlines the webpage’s structure for tasks like locating buttons for clicks. On a dynamic webpage, the DOM can frequently change, posing difficulties for the AI agent.
Beyond issues of reliability and integration, analysts also highlighted potential security concerns linked to entrusting AI agents with browser-level autonomy.
“These encompass managing authenticated browser sessions, engaging with unverified external websites, and ensuring that automated actions do not inadvertently submit incorrect or sensitive data,” Satapathy stated. “In regulated environments, this could complicate auditing and compliance reviews.”
