Google is moving its AI from an assistant that answers questions to an agent that completes tasks, even while you sleep.
Back
Google rolls out 3 new AI tools to automate online shopping
Google unveiled Gemini Spark, a personal AI agent designed to work 24/7 across its apps, part of a new agentic strategy aiming to automate user tasks and challenge the AI offerings from Microsoft and OpenAI.
"We are in that part of the cycle where people want to see real value in the products they use on a day-to-day basis," Sundar Pichai, CEO of Google and Alphabet, said during a press briefing ahead of the Google I/O 2026 keynote.
Gemini Spark runs persistently in the cloud on the new Gemini 3.5 Flash model and can execute multi-step tasks across Gmail, Docs, and third-party apps. It will be available to Google AI Ultra subscribers, with a restructured plan now starting at $100 per month.
The move pits Google (GOOGL) directly against OpenAI's ChatGPT agent and Microsoft's Copilot Cowork, signaling a major industry shift from conversational AI to autonomous agents that could reshape productivity and e-commerce.
Google Debuts Agent Payments Protocol With 'Teenager' Safeguards
A core part of Google's new commerce strategy is the Agent Payments Protocol (AP2), a system designed to let AI agents make secure purchases within user-defined boundaries. At launch, Gemini Spark will require explicit user approval for any transaction. However, the AP2 infrastructure is built for a more autonomous future where an agent could automatically complete a purchase if it meets criteria set by the user, such as specific brands, products, and spending limits.
Josh Woodward, VP of Google Labs, compared the spending guardrails to giving a teenager their first debit card. The system is underpinned by the Universal Commerce Protocol (UCP), an open-source standard Google announced this year that now includes Amazon, Meta, Microsoft, Salesforce, and Stripe on its tech council. Google also introduced the Universal Cart, which works across Google services to track price drops and surface deals.
Can Gemini 3.5 Flash Cut Enterprise AI Costs by $1 Billion?
The new agentic features are powered by Gemini 3.5 Flash, a model Google claims outperforms its previous frontier model, Gemini 3.1 Pro, while running four times faster than comparable models. The speed and cost-efficiency are central to Google's pitch for agentic AI.
Pichai argued that large enterprise customers processing roughly one trillion tokens per day could save over $1 billion annually by shifting most of their workloads to a mix of Flash and Pro models. This cost argument is critical in a market where CIOs are reportedly exhausting their annual token budgets in a matter of months. Internally, Google's use of the model has grown from half a trillion to over three trillion tokens per day since March, providing a massive feedback loop for improvement.
The announcement lands in a fiercely competitive environment. OpenAI, Anthropic, and Microsoft are all racing to build their own autonomous agents, each with a different architectural approach. While Anthropic's agent works on a user's desktop and Microsoft's is tied to Office 365, Google's bet is on a cloud-persistent agent deeply integrated with its own ecosystem of services. The success of Gemini Spark will depend not just on its technical capability, but on whether users are willing to trust an AI agent with a deeper level of autonomy and access to their digital lives.
This article is for informational purposes only and does not constitute investment advice.
[1] Google expected to court coders, consumers at I/O conference[2] How to use Google’s new AI agents to go beyond your standard searches[3] Google’s new AI agent can draft your emails, monitor your inbox and eventually spend your money[4] Google I/O 2026: Will AI music tools like Google Flow Music make creativity easier or originality weaker?
