Gemini Enterprise Rooms: Smart Playbooks, Surprising Goals

Gemini Enterprise may be about to become a very different product. On 31 August 2026, TestingCatalog reported that Google is prototyping a feature called Rooms inside Gemini Enterprise — shared workspaces where a team collaborates with colleagues and with Gemini around one specific objective. Creating a Room involves defining three things: a goal, a playbook describing how Gemini should operate, and a knowledge base of connected files. It is an unannounced prototype found in recent builds, and Google has not indicated any plan to ship it publicly.

That caveat matters, but so does the direction of travel. Rooms is the clearest signal yet that Google wants Gemini Enterprise to stop behaving like an assistant layered over company data and start behaving like project management and operational collaboration software — a place where artificial intelligence is a persistent team asset rather than a private chat window. For any organisation shaping an AI strategy around agentic tools, that shift deserves attention now, not when the feature ships.

This article walks through what the leak actually shows, how Goals and Playbooks work, how Rooms extends the Projects feature Google announced at Cloud Next 2026, how the concept compares with ChatGPT, Claude and Copilot equivalents, and what a sensible business does while the feature is still behind the curtain. Every figure is traceable to a source in the References section.

Table of contents

What the Gemini Enterprise Rooms Prototype Actually Is
Goals, Playbooks and Knowledge Bases: How a Room Is Built
From Projects to Rooms: The Gemini Enterprise Timeline
How Rooms Compares With ChatGPT, Claude and Copilot
What Rooms Says About Google’s Agentic Workplace Strategy
Gemini Enterprise Pricing and Where Rooms Could Land
What Your Business Should Do Before Rooms Ships
Open Questions Google Has Not Answered
Gemini Enterprise Rooms FAQ
References

What the Gemini Enterprise Rooms Prototype Actually Is

The discovery comes from TestingCatalog’s Alexey Shabanov, who found references to Rooms in recent Gemini Enterprise builds. The feature “remains marked as a prototype”, and the report is careful to say Google has not committed to a public release. No other outlet has independently confirmed the feature, so everything below describes a prototype that could change or disappear.

An “all-in-one place” built around one objective

Rooms are described as an “all-in-one place” environment where Gemini becomes an expert using connected files while keeping team conversations focused. The core idea is scoping: instead of one general assistant that knows a little about everything, a Room binds Gemini to a defined objective, a defined way of working and a defined set of sources. Team members and the model share the same context inside the same space.

People and Gemini in the same conversation

According to the report, teams can add members and files to a Room, maintain shared context, and communicate either with each other or with Gemini inside the same space. That is a meaningful departure from the dominant pattern of enterprise AI use, where each employee runs a private conversation and shares conclusions by copy-paste. In a Room, the conversation itself is the shared artefact.

What Rooms is not — yet

The report is explicit about the boundaries of the evidence. There is no confirmed launch window, no final name, and no list of supported data sources. TestingCatalog raises Google Meet integration — connecting call context to a Room’s decisions and tasks — as an obvious extension, while stressing that “there is currently no evidence that Meet integration is part of the prototype.” Treat every capability described here as provisional.

Goals, Playbooks and Knowledge Bases: How a Room Is Built

The most interesting detail in the leak is the setup flow. A Room is not just a folder with a chat attached; creating one involves defining a goal, a playbook describing how Gemini should operate, and a knowledge base. Those three elements are the difference between a workspace that stores context and a workspace that directs behaviour.

The goal: an explicit objective for the workspace

The goal gives the Room a stated purpose — ship the Q4 pricing review, run the customer migration, draft the tender response. Because the objective is explicit rather than implied by whatever files happen to be present, Gemini can evaluate its own contributions against it. That is the same design instinct behind goal-directed agents across the industry: an agent with a target produces measurably more useful output than an agent with a pile of documents.

The playbook: operating instructions for Gemini

The playbook describes how Gemini should operate inside the Room — the closest existing analogues are custom instructions in ChatGPT Projects or a Claude Project’s project-level guidance, but attached to a team space rather than an individual’s workspace. A playbook is where a team encodes its standards: tone, methodology, approval steps, what to always check, what never to do. Organisations that have invested in intelligent automation will recognise the pattern — the value is less in the model and more in the documented procedure the model follows.

The knowledge base: scoped, connected files

The third element is a knowledge base of connected files, which Gemini uses to “become an expert” for that Room. Scoping matters for quality — retrieval over a curated set beats retrieval over an entire drive — and it matters for governance, because a Room’s membership and sources define exactly who and what informs the model’s answers.

Room element	What it defines	Closest existing analogue
Goal	The objective the whole workspace is working towards	Project brief in work-management tools; no direct AI-native equivalent
Playbook	How Gemini should operate — methods, standards, constraints	Custom instructions in ChatGPT Projects; Claude Project instructions
Knowledge base	The connected files and sources Gemini treats as its expertise	Project files in ChatGPT and Claude; NotebookLM sources
Members	The colleagues who share the Room’s context and conversations	Shared projects (ChatGPT Business); Copilot Pages collaborators

From Projects to Rooms: The Gemini Enterprise Timeline

Rooms did not appear from nowhere. It is the next step in a progression Google has been running since Gemini Enterprise launched, and the prototype record shows the Goals idea has been visible since spring.

October 2025: the “front door” launch

Google Cloud CEO Thomas Kurian introduced Gemini Enterprise on 9 October 2025 as “the new front door for AI in the workplace”, bundling six components: Gemini models, a no-code workbench, pre-built Google agents, secure connections to company data across Google Workspace, Microsoft 365, Salesforce and SAP, central governance, and a partner ecosystem Google put at more than 100,000 companies. The launch post claimed nine of the top ten AI labs run on Google Cloud and that 65% of Google Cloud customers use its AI products.

March 2026: Projects spotted in builds

By 2 March 2026, TestingCatalog had already found a Projects feature in Gemini Enterprise builds — chats grouped by topic, file uploads, and “specific goals that guide how Gemini responds”. The goal-directed workspace concept, in other words, has been in Google’s prototype pipeline for at least six months. The same report noted Gemini 3.1 Pro entering the platform’s model selector shortly after its 19 February 2026 release — our Gemini 3.1 Pro guide covers that model in detail.

April 2026: Projects announced at Cloud Next

At Cloud Next 2026 in Las Vegas — 32,000 attendees, 260 announcements — Google made Projects official: “a shared workspace for teams and agents… A dynamic, highly collaborative workspace where human experts and agents actively co-create side-by-side”, unifying context across Google Workspace, Microsoft OneDrive and team chats. The same wave added Inbox, Skills, Canvas, Agent Designer and an Agent Gallery. Google’s framing was blunt: “Real work is a team sport. We are moving AI from a solo conversation to a persistent, shared team asset.”

August 2026: Rooms goes further

Rooms, per the new report, takes the Projects idea further “by giving the workspace an explicit objective, operating instructions, and a more structured knowledge layer”. Alongside it, TestingCatalog spotted the Customization area being reorganised into dedicated Agents and Memory sections — consistent with the platform’s earlier moves on skills, covered in our piece on the Gemini skills management section.

Date	Milestone	Status
9 Oct 2025	Gemini Enterprise launches as the “front door for AI in the workplace”	Shipped
2 Mar 2026	Projects prototype spotted in builds, with goals guiding responses	Leak, later confirmed
22–23 Apr 2026	Projects, Canvas, Inbox, Agent Designer announced at Cloud Next 2026	Announced
Jul 2026	Agent platform demos: long-running agents, BYO-MCP, connectors	Shipped
31 Aug 2026	Rooms prototype with goals, playbooks and knowledge bases reported	Internal prototype

The cadence is worth reading as a chart. Counting months from the October 2025 general-availability launch: the Projects prototype surfaced about 5 months in, the Cloud Next announcements landed at 6 months, the agent platform demos at roughly 9 months, and the Rooms leak at 10 months.

Months from Gemini Enterprise GA to each collaboration milestone

Projects prototype spotted 5 months

Projects announced at Cloud Next 6 months

Agent platform demos 9 months

Rooms prototype reported 10 months

How Rooms Compares With ChatGPT, Claude and Copilot

Every major AI vendor has converged on some version of the persistent, shareable workspace. The differences are in what the workspace actually encodes.

The competitive landscape in one view

OpenAI’s ChatGPT Projects give individuals a container of files, custom instructions and conversation history, and shared projects extend that container to teams on Business, Enterprise and Edu plans. Anthropic’s Claude Projects, launched in June 2024, offer 200K-context workspaces with custom instructions and an activity feed for sharing work. Microsoft’s Copilot Pages, announced 16 September 2024, is “a dynamic, persistent canvas… designed for multiplayer AI collaboration” inside the Microsoft 365 stack.

Capability	Rooms (prototype)	ChatGPT shared projects	Claude Projects	Copilot Pages
Explicit goal for the space	Yes — defined at creation	No — implied by content	No — implied by content	No — page-level content
Operating instructions	Playbook per Room	Custom instructions	Project instructions	No dedicated layer
Scoped knowledge base	Connected files per Room	Project files	Project knowledge, 200K context	Grounded in Microsoft 365 data
Team conversation with the AI	Members chat together or with Gemini	Shared files, instructions and history	Shared via activity feed	Multiplayer editing on a canvas
Availability	Internal prototype, unannounced	Shipped on team plans	Shipped since June 2024	Shipped since late 2024

Where the Rooms design is genuinely different

The structural difference is that a Room makes objective and method first-class configuration rather than content. In ChatGPT or Claude, a disciplined team can paste a brief and working rules into instructions; in Rooms, the goal and the playbook are the schema of the workspace itself. If Google ships it, the differentiator will be that a Room can, in principle, be evaluated — did the workspace achieve its stated goal, and did the model follow its playbook? That auditability is exactly what enterprise buyers keep asking agent vendors for.

What Rooms Says About Google's Agentic Workplace Strategy

TestingCatalog’s own reading is that Google has been “steadily repositioning Gemini Enterprise as an agentic workplace platform, with long-running agents, Projects, Canvas, an Agent Gallery, enterprise connectors, and persistent agent memory all forming parts of that strategy.” Rooms fits that arc precisely.

From assistant to operating layer

A chat assistant answers questions; a platform hosts work. Goals and playbooks are the vocabulary of work management — objectives, procedures, deliverables — grafted onto an AI product. The prototype moves Gemini Enterprise into territory occupied by project management software, and it does so with the model in the room rather than bolted on. The reorganised Agents and Memory sections point the same way: agents you can inspect and configure, and memory you can manage, are platform features, not chat features.

Governance is the quiet selling point

For enterprise cybersecurity and compliance teams, the Room construct is attractive for an unglamorous reason: scope. A workspace whose members, sources, objective and operating rules are all declared is far easier to audit than a thousand private chats. Pair a playbook with the platform’s central governance controls and you get something an IT governance function can actually review — a documented statement of what the AI is for and how it is allowed to behave, alongside the access controls a cybersecurity review would demand anyway.

Gemini Enterprise Pricing and Where Rooms Could Land

Rooms has no price, tier or date — it is an internal prototype. But the platform it would land in has a published price list, which is where any budgeting conversation starts.

The current price list

Google’s product page lists Gemini Enterprise Business at $21 per seat per month and Gemini Enterprise Standard at $30 per seat per month, with a Plus tier on contact-sales terms, a consumption-based option with a $0 seat fee where you pay only for what you use, and 30-day trials. At those two published seat prices, Standard costs $9 more per seat per month than Business — a 43% premium on the $21 base.

Published Gemini Enterprise seat pricing, USD per seat per month

Consumption option seat fee $0

Business $21

Standard $30

Reading the tiers for a Rooms future

Collaboration features tend to be the levers vendors use to differentiate team tiers, and a feature whose entire premise is shared context would be a strange fit for a single-seat plan. If Rooms ships, expect it to matter most on the tiers organisations already buy for teams — and expect the consumption option to make Room-hosted agent activity a metered cost worth modelling before rollout.

What Your Business Should Do Before Rooms Ships

A prototype leak is not a procurement trigger. It is, however, a free look at where the platform is going, and the preparation it rewards is preparation that pays off regardless of whether Rooms ever ships.

Write the playbooks now

The hard part of a playbook is not typing it into Gemini Enterprise; it is agreeing what it should say. Teams that document how work should be done — checks, standards, escalation rules, tone — can drop that material into any AI workspace the day it arrives, whether that is Rooms, ChatGPT Projects or a Copilot agent. Firms that treat business automation as a process-documentation discipline first and a tooling decision second consistently get more from every platform they adopt.

Curate the knowledge before the knowledge base

A Room’s answers will only be as good as its connected files. Auditing which documents are authoritative, current and safe to expose to a model is slow work, and it is the same work every retrieval-based tool needs. Starting it now converts a future Rooms rollout from a data-cleanup project into a configuration task.

Decide who owns goals and playbooks

A workspace with declared objectives and operating rules needs an owner for both. Decide now whether playbooks are set by team leads, reviewed by compliance, or templated centrally — the answer shapes training, governance and how many Rooms you would actually want. Our guide to the wider agent tooling market in the AI models and tools hub is a useful companion for mapping which platforms your teams already touch.

Open Questions Google Has Not Answered

The leak leaves the important commercial questions open, and it is worth being precise about what is unknown.

Will Rooms replace Projects or sit beside it?

TestingCatalog flags this directly: whether Rooms will replace or coexist with Projects is unresolved. The two occupy overlapping ground — Projects already gives teams shared, goal-aware context — so a merged future is as plausible as a two-feature lineup. Buyers who anchor plans on Projects today should assume the construct, not the name, is what persists.

Which sources, which tiers, which timeline?

Supported data sources are unconfirmed, the name is provisional, and there is no availability signal at all. The Meet integration idea is explicitly speculation. None of the platform’s published pricing mentions Rooms, and no Cloud Next-style announcement has referenced it. Anything more specific than “Google is experimenting here” outruns the evidence.

Will enterprises actually work this way?

The deeper unknown is behavioural. Rooms bets that teams want to meet their AI in a shared, structured space rather than each keeping a private assistant. Shared projects and Copilot Pages point the same direction, but none of these vendors has yet published adoption numbers proving the multiplayer pattern sticks. The winner of that argument will define what enterprise AI software looks like from here.

Gemini Enterprise Rooms FAQ

Is Rooms available in Gemini Enterprise today?

No. Rooms is an internal prototype found in recent builds and reported on 31 August 2026. Google has not announced it, and there is no confirmed release plan, name or tier.

What are Goals and Playbooks in the Rooms prototype?

They are two of the three elements defined when a Room is created: the goal is the workspace’s explicit objective, and the playbook is the operating instructions describing how Gemini should behave in that Room. The third element is a knowledge base of connected files.

How is a Room different from a Gemini Enterprise Project?

Projects, announced at Cloud Next 2026, give teams and agents a shared workspace with unified context. Rooms — per the leak — adds an explicit objective, operating instructions and a more structured knowledge layer on top of that shared-workspace idea. Whether the two merge is an open question.

What does Gemini Enterprise cost?

Published pricing is $21 per seat per month for Business and $30 per seat per month for Standard, with a contact-sales Plus tier, a consumption option with a $0 seat fee, and 30-day trials. Rooms has no published pricing because it has not been announced.