Gemini Agent for Sales Teams: Google's Universal AI Coworker
At Gemini at Work 2026 on October 8, Google Cloud CEO Thomas Kurian announced the Gemini agent — "a single, universal agent for work that answers your questions, handles your knowledge work, creates your images and media, and writes and runs code." One agent, every surface: web, mobile, desktop, even a command line. It's in private beta now, with general availability expected around the end of October or early November.
Buried under the platform language is the detail that matters for revenue teams: Google is shipping coworker agents — persistent AI teammates that get their own @agents.company.com email address, their own calendar, their own entry in the company directory. You don't prompt them. You email them, schedule them, and assign them objectives like you would a new hire.
And in a week where every vendor is metering agents by the action, the conversation, or the credit, Google said the quiet part: it does not plan per-seat licensing for coworker agents in Workspace.
Here's what launched, how it compares to the agents Salesforce, OpenAI, and HubSpot shipped in the last month, and what a sales team should actually do about it.
What Google Actually Shipped
The Gemini agent collapses Google's scattered enterprise AI surface — Gemini Enterprise, Workspace AI, Agentspace — into one agent you interact with four ways:
- Chat with it like any assistant
- Assign it objectives and let it plan the work itself
- Schedule it to run recurring jobs
- Trigger it from events — a form submission, an inbound email, a CRM change
The architecture is the real shift. The agent runs persistently in the cloud with shared memory and a personalization graph, so a multi-day job keeps running after you log out. It spawns temporary sub-agents for individual tasks and reports back when done. If you've read our coverage of OpenAI's Dots always-on agents, the pattern is familiar — the industry has converged on "agents that never clock out."
It reaches beyond Google's walls, too. The agent works across Google Workspace, Microsoft 365, and Slack, and connects to enterprise systems including Salesforce, ServiceNow, Jira, Snowflake, and Databricks. For grounding, Google announced a Knowledge Catalog for business definitions — during development, Bloomberg Media reported a 63% lift in SQL query accuracy when queries were grounded in it.
One more surprise: Smart Routing sends tasks across both Gemini and Anthropic's Claude models. Google is routing work to a competitor's model when it's the better fit for the task. That's a pragmatic concession to how enterprises already operate — most teams we talk to run Claude for sales workflows alongside Gemini already.
Coworker Agents: The Feature That Matters for Sales
Every vendor now sells "digital workers." Salesforce has Agentforce Coworker and the Piper and Hunter SDR agents. OpenAI has Dots. What Google added is an identity model the others don't have:
A coworker agent gets a real Google Workspace account — email address, calendar, storage, directory entry. That sounds mundane. It isn't:
- Reps interact with it like a teammate. Forward a prospect thread to
research@agents.company.comand get an account brief back. CC it on a deal thread and it keeps the CRM updated. No new interface to learn — the interface is email, which is where sales lives anyway. - It can be invited to things. Put the pipeline-review agent on the Monday call invite. Schedule the follow-up agent to sweep stalled deals every Friday.
- IT can govern it like an employee. Provisioning, permissions, and offboarding run through standard identity management. Every action lands in the audit log attributed to the agent, not to whichever human triggered it. Agent identity is cryptographically attested, permissions propagate via OAuth, and an Agent Gateway acts as what Google calls an "AI network firewall."
That last point is why this launch will clear security review faster than most. The number one blocker we see when sales teams try to deploy AI SDRs is not capability — it's that security teams can't answer "who did that?" when an agent touches the CRM. A directory-native agent with its own audit trail answers it.
Sales Workflows This Unlocks
Concrete plays a Workspace-based sales org can run once this hits GA:
1. An inbound triage agent with its own inbox. Route demo-requests@ to a coworker agent that enriches the lead, checks ICP fit against your Knowledge Catalog definitions, books the meeting, and logs it to Salesforce via the connector. This is the same job Salesforce's Piper does for $2 a conversation — except here it lives in the Gmail stack your team already pays for.
2. Scheduled pipeline hygiene. A recurring agent job that sweeps the CRM every morning for deals with no next step, stale close dates, or ghosted threads, then emails each rep their fix list. Multi-day persistence means it can chase its own follow-ups.
3. Event-triggered account research. New opportunity created in Salesforce → agent builds a brief from the account's website, news, and your internal docs → drops it in the deal folder and calendar invite before the first call. Pair it with the Gemini Skills your team has already written — skills define the playbook, the agent executes it unattended.
4. RFP and security questionnaire grinding. The Commerzbank case Google cited — manual document review cut from 20 hours to one — is the same shape as answering a 200-row security questionnaire from your champion's procurement team.
The common thread: these are jobs sales ops currently does with a patchwork of Zapier, spreadsheets, and interns. An agent with an email address and OAuth-scoped CRM access is a cleaner primitive.
Pricing: The Wedge Google Is Driving
Google did not announce pricing, and final metering models are undisclosed. But the statement that there's no per-seat licensing planned for coworker agents in Workspace is aimed directly at how everyone else charges:
| Vendor | Agent pricing model |
|---|---|
| Salesforce Agentforce | $0.10 per action or $2 per conversation |
| OpenAI Dots | Bundled at $100+/user Pro tier, 15-task/30-runs-per-hour limits |
| HubSpot Agent Hub | Credit-metered — ~100 credits per lead worked |
| Google coworker agents | No per-seat plan; metering TBD |
Treat "no per-seat" with appropriate skepticism — Google will meter something, likely tokens or compute, and undisclosed pricing is a reason to pilot, not to commit. But the strategic intent is clear. Google claims 8 million paid Gemini Enterprise seats across 4,200 companies and nearly 90% of the Fortune 100; it can afford to make agents a retention feature for Workspace rather than a new line item. Per-project spend caps with real-time pause are built in, which is more cost governance than most agent platforms launched with.
How It Stacks Up Against the Last 30 Days of Agent Launches
October has been absurd. In four weeks we've covered Salesforce's Koa model and AIforce suite, Gong, Apollo, and ZoomInfo all shipping agent builders in the same week, Claudeforce putting Salesforce inside Claude, and OpenAI's Dots. Where Gemini agent fits:
- vs. Salesforce: Agentforce agents are strongest inside the CRM with CRM data. Gemini's coworker agents are strongest around the CRM — email, calendar, docs, research — with Salesforce as one connector among many. If your reps live in Gmail and only visit Salesforce under duress, Google's shape fits better.
- vs. Microsoft Copilot: Microsoft has native-integration advantage in Microsoft-standardized orgs and existing enterprise agreements. Google's counter is cross-application context — one agent spanning Workspace and Microsoft 365 and Slack — plus multi-model routing. If your company runs Office, wait for evidence it works well there.
- vs. point-solution AI SDRs: Vertical sales agents still win on out-of-the-box sales judgment — a universal agent doesn't know what a good discovery question sounds like. Google's industry-specific agents are starting with financial services and legal (in preview); there's no sales-specialized agent yet.
- vs. the model layer: This launch is about orchestration, not model quality. Google's frontier-model story for sales is Gemini 4 Argon, and Smart Routing means the agent may quietly use Claude for some of your tasks anyway.
What Sales Leaders Should Do Now
- If you're on Workspace: ask your Google rep for the private beta. The identity and audit model means security review can start now, before GA. Inventory the three workflows above and pick one.
- If you're on Microsoft: nothing to do yet. The cross-suite claims need proof, and Copilot's agent roadmap is the default path for your stack.
- If you're mid-evaluation on an AI SDR vendor: finish the evaluation, but add a column for "does Workspace eat this feature by Q1?" Inbound triage and meeting booking are squarely in the blast radius.
- Watch the metering announcement. "No per-seat" is a direction, not a price. The number that matters is cost per worked lead, and nobody can compute it yet.
The pattern across this entire launch wave — Salesforce, OpenAI, HubSpot, now Google — is that always-on agents with real identity are becoming infrastructure, not features. The teams winning with them aren't the ones with the fanciest agent; they're the ones feeding agents clean signal about which accounts are actually in-market. That's the layer we build at MarketBetter: visitor identification and intent signals that tell your agents — Gemini, Agentforce, or ours — who to work and why. Book a demo to see how signal-fed agents outperform prompt-fed ones.
FAQ
What is the Gemini agent? A single universal AI agent for work announced by Google Cloud on October 8, 2026 at Gemini at Work. It answers questions, handles knowledge work, creates media, writes and runs code, and executes multi-day tasks persistently in the cloud across web, mobile, desktop, and CLI.
What are Gemini coworker agents? Persistent agents with their own Google Workspace accounts — email address, calendar, storage, and directory entry. Teams interact with them by email or meeting invite, and IT governs them through standard identity management with per-agent audit trails.
When is the Gemini agent available? It's in private beta as of October 8, 2026, with general availability expected around the end of October or early November 2026. Industry-specific agents for financial services and legal are in preview.
How much does the Gemini agent cost? Google has not announced pricing. A company representative said there are no plans for per-seat licensing of coworker agents in Workspace, but the final metering model is undisclosed.
Does the Gemini agent work with Salesforce? Yes. Connectors cover Salesforce, ServiceNow, Jira, Snowflake, Databricks, Confluence, and Microsoft 365/Teams, with OAuth-propagated permissions and action-level audit logging.