AI Assistants Set to Automate Online Tasks in the Emerging “Agentic Web” Era
The digital landscape is undergoing a major transformation as leading tech companies like Google and Microsoft accelerate the development and deployment of AI agents designed to autonomously manage users’ online tasks. This emerging “agentic web” signals a shift from traditional browsing and manual interactions to AI-driven automation and delegation.
Google’s Project Mariner is rapidly evolving from a research initiative into a core service component. Integrated into the Gemini API, these agentic capabilities will soon be accessible in widely used consumer products like Chrome, Search, and the Gemini app via a new “Agent Mode.” This functionality enables AI to perform multi-step tasks, such as refining apartment searches on Zillow, adjusting filters, and even scheduling viewing appointments. Similarly, Microsoft’s Build 2025 conference emphasized a strategy centered on “AI agents,” including the introduction of an “Agent Store” within Microsoft 365 Copilot, allowing users to explore and deploy specialized agents for various tasks.
This transition to AI-driven task automation represents a fundamental shift in user interactions with online services. Instead of navigating complex interfaces or juggling multiple apps for tasks like booking travel, managing appointments, or conducting research, users will increasingly rely on AI agents to handle these processes. These agents will interpret user intent, interact with relevant services, and complete tasks with minimal intervention.
While the agentic web promises unprecedented convenience and efficiency—enabling AI to book flights, arrange transportation, and manage schedules from a single request—it also raises critical concerns. As AI agents gain autonomy, issues surrounding data privacy, security, and user control come into focus. Consumers will need confidence that these systems act in their best interest and handle personal data responsibly. Additionally, errors or unintended consequences could arise if agents misinterpret instructions or encounter unexpected scenarios online.
Microsoft’s vision for an “open agentic web,” where AI agents interact across various platforms, further highlights both opportunities and challenges. Ensuring interoperability, security, and accountability will be essential as this new digital paradigm unfolds. For users, the emergence of the agentic web offers a glimpse into a more automated and personalized future—one that demands careful navigation and a deeper understanding of how intelligent systems operate.
To provide insight into the evolving AI landscape, the following table summarizes key developments announced by major players in late May 2025:
AI Landscape Snapshot – Key Announcements (Late May 2025)
| Company | AI Model/Product | Key New Feature/Capability | Stated Consumer Benefit/Impact | Availability/Rollout |
| Gemini 2.5 Series, Deep Think | Enhanced reasoning, 50x token processing increase, personal file integration for Deep Research, Canvas integration | More personal, proactive, powerful AI assistance; dynamic content creation (infographics, quizzes, podcasts) | Deep Think (experimental for 2.5 Pro); Deep Research enhancements rolling out; Canvas integration coming soon 1 | |
| Google Beam | AI-first 3D video communications platform, realistic 3D experiences from 2D streams | More immersive and natural video calls, breaking down language barriers with real-time expressive translation | Beta for AI Pro/Ultra subscribers (English/Spanish translation), more languages soon 3 | |
| Project Astra / Gemini Live | Camera and screen-sharing capabilities for universal AI assistant | AI understands visual context from camera/screen for more interactive help; coming to Search Live | Integrating into Gemini Live; Search Live in Labs this summer 1 | |
| Project Mariner / Agent Mode | AI agents for web interaction and task completion (e.g., apartment hunting, booking tickets) | Automated task completion on the web, more done with less effort | Capabilities via Gemini API; Agent Mode in Gemini app (experimental for subscribers soon); coming to Chrome, Search 1 | |
| AI Mode in Search | End-to-end AI search for complex queries, deep research, live visual search, agentic capabilities, AI shopping | More thorough, conversational, and actionable search results; virtual try-on, agentic checkout | Rolling out in U.S. (opt-in via Labs); Deep Search, Search Live, agentic features in Labs; new shopping experience coming 1 | |
| Meta | Meta AI | Reached 1 billion MAU; focus on personalization, voice, entertainment; plans for paid recommendations/subscriptions | More personalized AI assistant; potential for premium features | 1 billion MAU currently; monetization plans for the future 4 |
| Anthropic | Claude Opus 4 | New model launched with proactive ASL-3 safety standards (e.g., harder to jailbreak for CBRN misuse) | Access to a powerful new model with enhanced, precautionary safety measures against misuse | Claude Opus 4 and ASL-3 protections active now 9 |
| xAI | Grok on Telegram | Integration of Grok chatbot into Telegram via $300M deal | Grok accessible to Telegram’s 1B+ users; potential for new AI assistant choice within the app | Integration following the deal announcement 12 |
| Microsoft | Agent Store (M365 Copilot) | Marketplace to build, publish, and discover AI agents for Microsoft 365 | Access to a growing library of specialized AI agents for diverse tasks, boosting productivity | Announced at Microsoft Build 2025 3 |
| DeepSeek | DeepSeek-R1-0528 | Upgraded reasoning model with performance parity to OpenAI o3 & Gemini 2.5 Pro on benchmarks, efficient architecture | Access to a highly capable open-source reasoning model, potentially at lower computational cost | Available on Hugging Face now 17 |
