Google is rolling out a version of Gemini that does not just answer questions — it takes on work. The company has given its AI agent capabilities to plan multi-step tasks, hand off pieces of work to other specialized AI "subagents," and operate with its own identity inside a business, complete with a company email address.
The rollout is starting with business customers rather than consumers, a sequencing choice that signals where Google sees the near-term money. Instead of a chatbot that waits for a prompt and returns an answer, this version of Gemini can be assigned a goal, break it into steps, pull in different AI models suited to each step, and act across connected business systems without a human re-prompting it at every stage.
The workplace identity piece is the detail worth sitting with. Giving an AI agent its own email address means it can be added to threads, receive assignments, and show up in a company directory much like a contractor or new hire would. That is a meaningful shift from AI as a tool you open to AI as an entity you provision, manage, and eventually have to account for in your systems.
Google joins a field that has been moving in this direction for months. Microsoft has pushed agent features through Copilot, OpenAI has built agent capabilities into ChatGPT for business tiers, and a wave of smaller startups have pitched "AI employees" as their core product. What distinguishes Google's move is the depth of integration with its existing business software — Gmail, Docs, Sheets, and Workspace generally — rather than a bolt-on feature.
Why it matters
The broader pattern here is a race among the major AI vendors to move past the chatbot interface entirely. Every large AI company now has some version of an agent pitch: give it a goal, let it work, check in only when it needs a decision. Google's addition of a workplace identity pushes that pattern a step further by treating the agent less like software and more like staff.
This also fits a recurring rollout sequence in enterprise AI: features land with paying business customers first, get refined under real usage and liability pressure, and migrate to consumer tiers later — often with a price tag attached once the capability proves its value internally.
What this means for small businesses
For a small business, an AI agent with email access and the ability to act across apps raises practical questions before it raises productivity. Who approves what it sends on the company's behalf? What happens when it makes a judgment call on a vendor negotiation or a customer email that turns out wrong? Access and permissions will need the same scrutiny you'd give a new employee's login credentials — arguably more, since the agent can work at a pace no human assistant can.
There is real potential upside in offloading repetitive, multi-step admin work — scheduling, data entry across systems, drafting and routing documents. But the cost structure for agentic features has consistently trended toward premium pricing once the free preview period ends, based on how other AI vendors have handled similar launches. Budget for that before building workflows around it.
What to watch
Watch for Google's pricing details once the preview period ends, and watch how it handles permission boundaries — specifically whether business owners get granular control over what the agent can send or approve autonomously versus what requires human sign-off.
The bottom line
Gemini's shift toward an autonomous, identity-bearing agent is part of a broader industry move past chatbots and into AI systems that act on your behalf — small business owners evaluating it should treat onboarding an AI agent with the same access controls and oversight they'd apply to a new hire.