Analyst(s): Nick Patience
Publication Date: October 9, 2026
Google Cloud used its Gemini at Work 2026 event to launch the Gemini agent, a single agent for work that runs across Google Workspace, Microsoft 365, and Slack, routes jobs between Gemini and Anthropic Claude models, and gives coworker agents their own identities. We think Google’s decision to separate the agent from the model matters more than any individual feature.
What Is Covered in This Article:
- Google Cloud’s launch of the Gemini agent at Gemini at Work 2026, a single agent for knowledge work, content creation, and coding, a year after Google pitched Gemini Enterprise as the front door for AI in the workplace.
- Why Google’s decision to route work across Gemini and Anthropic Claude models matters more than any single feature in the launch.
- ETR data on enterprise use of Gemini and Claude model families and of the Gemini Enterprise Agent Platform.
- Coworker agents with their own identities, mailboxes, and calendars, and what they mean for governance and per-seat software pricing.
- The competitive implications for Microsoft, Salesforce, OpenAI, and Anthropic.
The News: Google Cloud announced the Gemini agent at its Gemini at Work 2026 event, describing it as a single, universal agent for work. Users assign it objectives rather than step-by-step instructions, and it answers questions, handles knowledge work, generates images and media, and writes and runs code from one interface and one API. The agent runs in the cloud with persistent memory and context, can be reached from web, mobile, desktop, the command line, Google Workspace, Microsoft 365, and Slack, and can operate headlessly inside third-party applications. It can spin up temporary sub-agents for multi-step jobs that run for hours or days, and it can create persistent coworker agents that receive their own Workspace accounts, including email addresses, calendars, and Drive storage.
Google says the agent routes each job to the model that fits best, drawing on its own Gemini models and Anthropic’s Claude models today, with OpenAI and open-weights models to follow. Governance features include cryptographically attested agent identities, role-based permissions, audit trails attributed to the agent, an Agent Sandbox with its own network boundary, and Agent Gateway, which enforces policy on all traffic in, out of, and between agents. Cost controls include smart routing between models and real-time spend caps set per project, with per-user caps to follow. Google also introduced industry specializations for financial services and legal, both in preview. The Gemini agent is in private preview, and Google expects general availability by the end of October 2026, starting in North America.
Google Gemini Agent Puts the Agent Ahead of the Model
Analyst Take: Enterprise agents have multiplied faster than most organizations can govern them through 2026, and the Google Gemini agent is Google Cloud’s attempt to collapse that sprawl into one agent with one API, one memory, and one set of controls that follows employees across Workspace, Microsoft 365, Slack, and the command line. It is the most coherent single-agent proposition a hyperscaler has put forward to date, although a good deal of what makes it interesting depends on rival platforms cooperating.
Google Cloud described Gemini Enterprise as the new front door for AI in the workplace when it launched the product at Gemini at Work in October 2025, and twelve months later, the company is making the same single-entry-point pitch. What sits behind the door has changed a great deal, however. Gemini Enterprise arrived as a chat interface that brought Gemini models, Agentspace technology, and prebuilt and third-party agents together for employees at $30 per seat per month, with the user still doing most of the steering. The Gemini agent takes objectives rather than prompts, works for hours or days without supervision, chooses between Gemini and Claude models, and creates coworker agents with identities of their own, all included at no additional charge where Gemini Enterprise is available.
Agentspace, Gemini Enterprise, and now the Gemini agent amount to three positions in less than two years, which shows how quickly the market’s idea of an enterprise AI front door has moved from somewhere to ask questions to somewhere to hand off work, and how much rebranding Google’s customers have had to absorb along the way. Google managed to acknowledge as much and says consumer and enterprise Gemini will run on the same backend within three months, which should help.
Google’s decision to treat the model underneath the Gemini agent as a separate, swappable choice is significant. The agent routes each job across Gemini models and Anthropic’s Claude models today, with OpenAI and open-weights models to follow, and Google argues that matching the model to the task raises accuracy on hard work and cuts cost on simple work. Administrators decide which models the organization can use, and users either pick one per task or leave it to an auto mode that starts work on a Flash model and escalates to a frontier model when the job gets harder. Google says only the top 10% to 15% of a typical workflow needs a frontier model, which it claims cuts costs by about a third. That is a sound engineering argument, but it is also a notable commercial concession from a company that has spent the best part of three years telling customers that Gemini models are the reason to buy Google Cloud AI.
Why Claude Sits Inside the Google Gemini Agent
Futurum’s September 2026 AI Product Series survey helps explain why Google would make that concession. Among 511 respondents asked about their plans for foundation model families in AI application development, 69% said they currently use Anthropic Claude and plan to continue, compared with 39% for Google Gemini. A further 18% said they plan to evaluate or adopt Gemini, so Google’s pipeline is healthy, but a universal agent that ran only on Gemini would have asked many enterprises to set aside the model family their developers already favor. Putting Claude inside the Google Gemini agent removes that objection and lets Google compete for the agent layer without first winning the model argument.
Memory, skills, tool connections, identity, and policy become sticky assets once the model is something the agent chooses on the customer’s behalf. Google makes the point itself, noting that the leading model changes every few months and that keeping the choice open means context, skills, and data stay where they are. We agree, and we expect every major platform vendor to make the same case in time. The risk for Google, however, is that it has now told customers the model is interchangeable while still needing them to believe that Gemini models and the TPUs underneath them are worth paying for.
Coworker Agents Raise Questions About Identity and Pricing
Coworker agents, which Gemini creates with a persistent role and their own Workspace accounts, including an email address on an agent’s subdomain, a calendar, Drive storage, and an entry in the company directory, are the features with the most consequences for enterprise IT. Colleagues work with them by adding them to a Chat space or tagging them in a document, and each carries a cryptographically attested identity with least-privilege permissions, with every action written to an audit trail attributed to the agent. Treating agents as governed, non-human identities is the right design, and Google has built it in from the start, which is more than can be said for many agent platforms that still borrow the permissions of whichever user launched them.
Agents with a directory entry, a mailbox, and a calendar look a lot like users, and most enterprise software is still priced per user. Google says it will not charge additional seats for the Gemini agent, keeping per-user SKUs for Gemini Enterprise and Workspace while billing advanced, long-running work against pooled quotas or on consumption, which shifts the real cost to tokens, sandbox compute, and the spend caps it has built into Cloud Billing. Google argues that consumption billing lets division heads set and track their own AI budgets rather than leaving IT to make one decision for everyone, which is a fair point, although it also turns AI cost management from a procurement exercise into an ongoing operational one. That model suits Google, which earns on consumption, far better than it suits application vendors whose seat counts could come under pressure as agents absorb tasks that once justified additional licenses. We expect the licensing of agent identities to become a live negotiating point in enterprise software renewals during 2027.
A Universal Agent Depends on Other Vendors’ Platforms
Google’s claim to universality rests partly on running inside Microsoft 365 and Slack, and here we are more skeptical. Microsoft has every reason to make Copilot the default agent in its own productivity suite and to control how third-party agents read and act on data in Outlook, Teams, and SharePoint, while Salesforce, which owns Slack, has its own Agentforce ambitions. Google’s list of supported tools is long, spanning Confluence, Jira, Salesforce, ServiceNow, its own BigQuery, Databricks, Postgres, Snowflake, and any Model Context Protocol (MCP) server, but the breadth of a connector list says little about how much an agent can actually do through each one, and the Gemini agent will work best inside Workspace, where Google controls the whole stack.
Google’s answer is to put the Gemini agent inside partners’ applications, including Salesforce, ServiceNow, SAP, and Slack, so those vendors keep control of their user experience and customer relationship. That is a shrewd pitch to SaaS vendors worried about agents stripping away their brand equity, and it sets Google apart, it argues, from rivals telling the market that SaaS is dead. It is a harder sell to Microsoft, which has a competing agent of its own, and it sits awkwardly alongside our point that coworker agents could erode the seat counts those same SaaS vendors depend on.
Workspace customers are the obvious first market, and Google has momentum to build on. Google is also making the Gemini agent the Gemini that Workspace users already see in the side panel of Docs, Sheets, and Slides, which gives it an installed base on day one that no standalone agent launch could match. The company says nearly 90% of the Fortune 100 use Gemini Enterprise, and Futurum’s September 2026 survey found that 22% of 469 respondents currently use the Gemini Enterprise Agent Platform for AI application development and plan to continue, with a further 21% planning to evaluate or adopt it. Turning that interest into a single agent that employees trust with multi-day, unsupervised work is a different matter, however. Google’s own customer examples, from Bradesco’s contract and fiscal reviews to DBS Bank’s chains of 70 to 80 specialized agents for corporate credit memos, are mostly bound, process-specific deployments with humans firmly in the loop.
Competitive Pressure on Microsoft, OpenAI, and Anthropic
The Google Gemini agent puts Google in direct competition with Microsoft’s Copilot and agent strategy, OpenAI’s enterprise push with ChatGPT, and, awkwardly, Anthropic’s own enterprise products, even as Claude models run underneath Google’s agent. Anthropic gains distribution and token revenue from the arrangement, but every enterprise that standardizes on the Gemini agent as its front door for delegated work is one that is less likely to standardize on Anthropic’s own interface. Futurum’s October 2026 TSIS survey gives Anthropic an overall Net Score of 78 (n=757), a measure of net spending intent that suggests Anthropic has plenty of momentum of its own and little need to settle for the role of model supplier to someone else’s agent. That tension between partner and competitor will shape how both companies price and position their agents over the next year.
Google has put forward the most complete enterprise agent design we have seen from a hyperscaler, with identity, policy, cost control, memory, and model choice designed in from the start. Whether it becomes the default way enterprises delegate work depends less on the architecture than on three things: how quickly a North America-first launch reaches other regions, how much access Microsoft and Salesforce allow on their platforms, and whether IT leaders are ready to let a single agent hold the memory of everything an employee does.
What to Watch:
- How quickly Google extends the Gemini agent beyond North America, particularly to European customers with data residency requirements.
- How Microsoft and Salesforce respond to a Google agent operating inside Microsoft 365 and Slack, including API access, data-access terms, and any limits on third-party agents.
- Whether Anthropic remains content as a model supplier inside Google’s agent or competes harder at the agent layer, and how quickly OpenAI models join Google’s routing roster, as Google says they will.
- How enterprises and software vendors license coworker agents that hold their own directory entries, mailboxes, and calendars, and whether per-seat pricing models adjust in response.
- Evidence that smart routing and real-time spend caps keep token costs under control at scale, given Google’s own acknowledgment that falling per-token prices have not stopped enterprise AI bills from rising.
- How Google and its customers measure return on the Gemini agent, given that the product tracks spend in detail but leaves productivity measurement to Google’s deployment teams and to customers themselves.
See the complete keynote announcement on the Gemini agent and Gemini at Work 2026 on the Google Cloud blog.
Disclosure: Futurum is a research and advisory firm that engages or has engaged in research, analysis, and advisory services with many technology companies, including those mentioned in this article. The author does not hold any equity positions with any company mentioned in this article.
Analysis and opinions expressed herein are specific to the analyst individually and data and other information that might have been provided for validation, not those of Futurum as a whole.
Other Insights From Futurum:
Google Returns to the Frontier With Gemini 4 Argon
Google’s Vertical AI Bet: Governance Matters More Than Models
Can Google Gemini Enterprise Unlock the Front Door for Business AI?
Author Information
Nick Patience is VP and Practice Lead for AI Platforms at The Futurum Group. Nick is a thought leader on AI development, deployment, and adoption - an area he has researched for 25 years. Before Futurum, Nick was a Managing Analyst with S&P Global Market Intelligence, responsible for 451 Research’s coverage of Data, AI, Analytics, Information Security, and Risk. Nick became part of S&P Global through its 2019 acquisition of 451 Research, a pioneering analyst firm that Nick co-founded in 1999. He is a sought-after speaker and advisor, known for his expertise in the drivers of AI adoption, industry use cases, and the infrastructure behind its development and deployment. Nick also spent three years as a product marketing lead at Recommind (now part of OpenText), a machine learning-driven eDiscovery software company. Nick is based in London.
