How Google Is Turning Search Into AI Agents

TL;DR

Google is transforming search from a tool that returns information into an agent that monitors events, browses websites, calls businesses, and helps complete real-world tasks. Gemini Spark runs continuously on Google Cloud, while a redesigned multimodal search box, prototype display glasses, and Gemini Omni extend AI assistance across search, navigation, apps, and media generation.

Transcript

AI, Google, Google. Google in search through and AI over Fore! Foreign! Foreign! AI agent agent Gemini era agentic payments agent agent introducing Gemini Spark. It's your personal AI agent that helps you navigate your digital life, taking action on your behalf and under your direction. It runs on dedicated virtual machines on Google Cloud and it's... Read More

Key Insights

• Google Search is evolving from a query-and-response product into an agentic system that can maintain conversations, monitor information, and take meaningful actions. Google described this transition as a foundation for changes across Search, Gemini, Chrome, Android, and other major products used by billions of people.
• Gemini Spark is a personal AI agent designed to act on a user's behalf and under the user's direction. It operates continuously on dedicated virtual machines in Google Cloud, allowing assigned work to continue around the clock even after the user closes a laptop.
• The redesigned Google search box accepts text, images, files, and videos, then reasons across those inputs. Google called it the largest upgrade to the search box since its debut more than 25 years ago and said the rollout was beginning at the time of the presentation.
• Information agents are designed to track changing real-world information instead of answering only once. A demonstrated agent periodically searched for restaurant and bakery tasting pop-ups in San Francisco, prioritized weekend events, notified the user about relevant discoveries, and continued monitoring every week.
• Agentic booking works by browsing multiple third-party websites to compare live availability and detailed requirements. In the karaoke example, AI Mode checked a Chicago Chinatown location, a specified date and time, capacity for nine people, food availability, and a two-hour session before presenting matching options.
• Google's calling capability is useful when businesses have incomplete online information. With permission, AI Mode can call providers, disclose that it is an automated Google service, ask about availability and costs, and summarize details such as base prices and additional charges for specific cleaning tasks.
• Google's prototype display glasses combine audio, a camera, and a display positioned in the right eye. Demonstrated functions included navigation, remembering a previously visited place, adding a coffee stop, placing a usual pickup order, showing visual cues, and accessing integrated productivity applications.
• Gemini Omni is presented as the next step in bringing Gemini intelligence to Google's media models. Google connected it to the broader goal of making Gemini genuinely multimodal, with the ability to accept any input, although the transcript does not provide detailed performance measurements or release information.

Questions & Answers

Q: How is Google turning Search into an AI agent?

Google is expanding Search from a system that returns information into one that can monitor developments, maintain an ongoing dialogue, browse websites, call businesses, and support real-world actions. Demonstrations included tracking weekend bakery events, finding karaoke rooms with several constraints, and calling cleaning providers. Users still direct the work and retain control when a final booking requires visiting a third-party site.

Q: What is Gemini Spark and how does it work?

Gemini Spark is described as a personal AI agent that helps users navigate their digital lives and takes actions on their behalf under their direction. It runs on dedicated virtual machines in Google Cloud and operates continuously, including after a laptop is closed. The presentation positions it as an always-available agent rather than a tool limited to an active chat session.

Q: What can Google's redesigned AI search box accept?

Google's redesigned search box can accept questions and materials across several modalities, including text, images, files, and videos. Search then reasons across those inputs instead of treating the box as a narrowly contained space. Google described the redesign as the biggest upgrade to its iconic search box since the box debuted more than 25 years ago, with rollout starting at the presentation.

Q: How do Google information agents monitor real-world updates?

Information agents periodically search for new information that matches a user's continuing interests and send notifications when relevant results appear. The demonstration created an agent for restaurant and bakery tasting pop-ups in San Francisco, especially on weekends. Example results included several bakery pop-ups and food or cultural tastings, and the system was set to continue checking every week for additional events.

Q: How does AI Mode find complex bookings online?

AI Mode uses live browsing to visit different websites, click and scroll through pages, check availability, and compare a request's specific constraints. For a private karaoke room, it evaluated Chicago Chinatown venues for a particular date, space for nine people, an evening time, a two-hour session, and food service. It presented matching choices and then linked users to third-party booking sites.

Q: Can Google AI call businesses for a customer?

Google AI can call businesses when online information does not adequately answer a customer's request, provided the customer grants permission. In the cleaning-service example, the system searched online first, then called several providers and reported their prices, availability, and extra charges. The automated call identified itself as a Google service and stated that calls could be monitored and recorded to improve services.

Q: What can Google's prototype display glasses do?

Google's prototype glasses provide audio, a camera, and a visual display in the right eye for assistance during everyday activities. Demonstrations referenced navigation to a place visited during the previous week, adding a coffee stop, ordering a usual drink through DoorDash, music with visual cues, AI camera functions, search, and integrations with Tasks, Keep, Calendar, and other workspace tools.

Q: What is Gemini Omni designed to accomplish?

Gemini Omni is described as the next step in Google's effort to bring Gemini intelligence into its media models and make Gemini genuinely multimodal. The stated direction is a model that can take any kind of input. The transcript introduces this goal but does not give specific benchmarks, pricing, availability dates, or a detailed technical comparison with other video-generation systems.

Summary & Key Takeaways

• Google presented search as an agentic service that can move beyond answering individual queries. Information agents periodically monitor real-world developments and notify users about relevant results, while booking tools browse multiple websites, compare availability and requirements, and direct users to third-party services where they retain control over the final transaction.

• Search can also call businesses when sufficient information or booking options are unavailable online. With customer permission, the system contacts providers, identifies itself as an automated Google service, asks about prices and availability, and reports the responses. Google said calling is already available in the United States for several service categories.

• Google also demonstrated prototype glasses designed for everyday assistance through audio, a right-eye display, cameras, navigation, visual cues, and integrations with apps such as Calendar, Tasks, and Keep. Gemini Omni was introduced as another step toward bringing Gemini intelligence into media models and accepting different forms of input through a multimodal system.