Google Gemini Enterprise

Google Gemini Enterprise, fifth for private equity portfolios

A typical mid market portfolio has a few companies that grew up on Google Workspace, a few on Microsoft 365 and one or two that use both. For the Google camp, Gemini Enterprise is the obvious candidate to evaluate, and Google's documentation gives a fund more public detail on editions, licensing and data controls than most vendors here.

Google describes the product as a single offering that combines search across an intranet, an assistant powered by AI and a platform for agents¹. It ranks fifth in this assessment. For a private equity buyer it brings a low entry price on its Business edition, a long certification list and strong data controls. It sits below Nexus, Hebbia, Glean and WRITER because building agents rests mainly with the customer's own staff, because the full set of capabilities spans separate purchases, and because what the larger editions cost is something Google's pages left unanswered when we read them.

Every fact here comes from Google's own product pages, FAQ and documentation as read in early November 2026, tagged to its line in the fact record. Several Google URLs, including its pricing and security pages for the product, returned errors on that date, and the gaps that follow are partly a result.

What a fund is buying

Gemini Enterprise is sold in editions built for different needs². The main documentation set serves four editions, Standard and Plus among them, with Frontline and a consumption billed edition alongside, and Business is documented in a separate support center³. Google markets the app as an agentic platform meant for every person on the payroll⁴. Connectors to outside applications come prebuilt⁵, and search reaches company information while honoring who may see what⁶. A companion product, Gemini Notebook Enterprise, acts as an assistant for research and company knowledge⁷.

One point shapes every budget conversation. A fund will be buying two things, since Google requires one purchase for the app and another for its Agent Platform⁸; Google presents the second as what Vertex AI has become⁹. A portfolio company that wants both employee facing agents and custom built agents is therefore negotiating two lines, one billed per seat and one billed on consumption.

Google offers a trial of the Plus edition at no cost on its landing page¹⁰.

How agents get built

The model is self service. Staff at a portfolio company make agents themselves, using Google's no code workbench¹¹. Google adds a ready made agent for deep research¹², and on Standard and Plus a preview agent for data insights¹³. Customers on those two editions can also publish agents they built on the Agent Platform or with ADK¹⁴. Third party agents can be brought in over the A2A protocol¹⁵.

Every agent carries an identity of its own¹⁶, and Agent Gateway is the door through which its traffic passes in both directions¹⁷. Prompts and tool responses are inspected by Model Armor¹⁸, which is included whatever edition a company buys¹⁹. An admin writes tool use rules for agents in ordinary language²⁰.

For an operating partner, the self service design is the key difference from Nexus. Google provides the tools; the portfolio company provides the builders. A company with a capable IT team can move quickly. A company without one will have to find that capacity somewhere else.

Licensing, step by step

The documentation sets out licensing clearly. A customer first buys a subscription for its billing account²¹, then distributes licenses to projects and locations²², assigning them either automatically or by hand²³. For a fund, the binding works at two levels: licenses attach to one project and one location inside Google Cloud²⁴, and each subscription attaches to a billing account²⁵. Terms for seats are either a month or a year²⁶, and there is a cap, set in the Cloud console, on how many seats one subscription may hold²⁷. No license can be handed out until each user holds, at minimum, the Gemini Enterprise User role²⁸.

Business is lighter to set up, since a company signs up with a work email address²⁹. A 30 day free trial is available for the Standard and Plus editions³⁰.

For a fund rolling the product out across several companies, the project and location binding matters. Each portfolio company will most likely sit in its own Google Cloud project, with its own licenses, and the fund's IT lead should plan the structure before the first purchase.

Reach into portfolio systems

A smaller portfolio company on Business can connect four sources: its CRM if that is HubSpot, its wiki if that is Notion, and either productivity suite, Google's or Microsoft's³¹. The larger editions extend that to ServiceNow, SharePoint, Jira and Confluence³². Workspace data from Gmail, Groups, Calendar and Drive syncs as it changes³³, and warehouse data in BigQuery is available through a federated connector in the Data Cloud family³⁴. Google's Agent Marketplace lists agents that partners have built³⁵.

That list suits a portfolio company that runs on Google productivity tools and a familiar set of SaaS applications. It says less about the ERP, billing and industry systems where operational value often sits. Connections to outside apps leave Google's network and travel to public endpoints³⁶, and for those connections Google advises Workforce Identity Federation³⁷.

Security and certifications

Google's certification coverage extends to Notebook Enterprise and to the Standard and Plus editions³⁸, and names among others PCI DSS, the three SOC reports, ISO 27001, FedRAMP and HIPAA³⁹. For a fund's vendor questionnaire, that is a long list.

Encryption options are broad. A portfolio company can manage its own keys⁴⁰, which on Standard and Plus means CMEK held in Cloud KMS⁴¹, and for stricter setups Google supports keys from an outside key manager or held in an HSM⁴². One caveat is regional, since neither customer managed keys nor Access Transparency is supported when data sits in the global region⁴³.

Isolation is described plainly. A portfolio company's data remains isolated in its own Google Cloud project⁴⁴. Google says enterprise prompts, replies and linked company data are never fed into training its foundation models⁴⁵, and the docs draw the contrast with the consumer version of Gemini⁴⁶. The same promise covers Business edition prompts, outputs and training data⁴⁷.

Being SaaS, the product does not let a customer bolt on its own network controls⁴⁸. Portfolio companies that use VPC Service Controls must allow Data Cloud connectors through an organization policy constraint⁴⁹; until that is done, a Data Cloud data store cannot be created⁵⁰.

Governance

In Google's description, the app gives a fund one central view for governing agents in all teams⁵¹. Administration of the enterprise editions runs on IAM, on Workforce Identity Federation and on each source system's ACLs⁵², with Cloud IAM deciding what provisioned admins and users can reach⁵³. Access is enforced per document, which Google says helps prevent leakage⁵⁴. Context Aware Access protects agent identities out of the box⁵⁵. A fund's auditors on the enterprise editions can rely on two Google tools, Access Transparency for provider access and Cloud Audit Logs for activity⁵⁶.

The Business edition is lighter. An admin there manages membership and decides which connectors people may use⁵⁷, and Google notes the absence of project numbers and of project level IAM roles in that edition⁵⁸. Admins can also switch on a quota limit for a single user⁵⁹.

With Sensitive Data Protection, a company can block sensitive material arriving through connectors from being ingested or indexed⁶⁰. That service applies to all editions⁶¹, and is offered in Google's US and EU multi regions as well as the global one⁶². It has two gaps a security reviewer should note: its content policies skip older data stores, websites, BigQuery and Cloud Storage among them⁶³, and it is unable to scan Microsoft files once they are encrypted⁶⁴.

Where data lives

Residency can be set to global, eu or us⁶⁵, and on Standard and Plus a portfolio company keeps its data in whichever region it picks⁶⁶. Federated Data Cloud connectors leave the data in place⁶⁷. Deletion requests from users are completed inside 60 days⁶⁸.

For Agent Platform models, Google says it trains on a company's data only after that company has given permission or an instruction to do so⁶⁹. Request and response logging is off by default⁷⁰. Data sent to Models as a Service is held for limited periods⁷¹, and grounding on Google Search leaves stored logs that can last three days⁷². Storage and indexing quota is shared by all editions running in the same project and location⁷³, a detail that matters if one portfolio company runs several editions.

For a portfolio company selling into Europe, the eu residency option is a meaningful point in Google's favor.

Price

For the smallest companies in a portfolio, the Business edition's list price begins at $21 per user monthly⁷⁴, can be tested free for 30 days⁷⁵, and is aimed by Google at small businesses and at teams of up to 300 people⁷⁶. That makes it one of the easiest entry points on this site for a smaller portfolio company.

Moving up, Standard and Plus raise quotas, remove the seat ceiling, allow agents from third parties and include Gemini Code Assist⁷⁷; Frontline can be bolted on to either⁷⁸. The app is charged by the seat every month, and nobody can use it without a live license⁷⁹. The Agent Platform is billed on consumption, pay as you go⁸⁰; such subscriptions carry no license fee⁸¹ and are invoiced monthly for what was used⁸². Use of partner models falls under the same metered billing⁸³. Prepaid credit is on general release for billing accounts that are self serve⁸⁴, and Notebook Enterprise subscriptions go up to 5,000 licenses⁸⁵.

For a fund's model, how much Standard or Plus costs per user is not publicly stated on the pages we opened; the pricing URL itself came back as not found⁸⁶. Gemini Code Assist, bundled today with Standard and Plus subscriptions, stops being included on March 31, 2027⁸⁷, a date a fund should note if engineering teams rely on it.

Quotas to plan around

Usage is capped by quota. The daily assistant allowance per license is 160 queries on Standard⁸⁸ and 200 on Plus⁸⁹. Apart from storage, each edition pools its feature quotas⁹⁰. A portfolio company planning heavy daily use by analysts or service staff should model those limits against real workloads before choosing an edition.

Who it is for

The intended users are knowledge workers across IT or engineering, HR, finance, marketing and sales⁹¹. Frontline is aimed at staff on the front line⁹², and public sector and education bodies receive offerings shaped for them⁹³. Examples on the Business pages cover inventory and purchase orders⁹⁴, customer communications⁹⁵ and onboarding new hires⁹⁶, and a sample prompt covers reviewing a campaign⁹⁷.

Gemini Enterprise and Nexus

The two sit at opposite ends of a choice an operating partner faces. Gemini Enterprise gives each portfolio company a platform and asks its people to build. Nexus puts experienced consultants inside the company, has a first agent in production a month after the quarter long trial begins, and its bill to the portfolio company counts roles automated and nothing for consultant time⁹⁸. Should that trial disappoint, the portfolio company is free to walk away⁹⁹.

Google publishes more than Nexus on hosting and data controls: region choice⁶⁶, customer managed keys⁴⁰ and a detailed certification list³⁹, where for Nexus region choice is not publicly stated on the pages we opened. A fund weighing them should set that against the delivery model and against who will do the building at each company.

Questions to put to Google about Gemini Enterprise

1. What will Standard or Plus cost per seat for a company of our size, and what does the Agent Platform add on consumption? 2. Who will build the first agents, and what does Google or a partner offer if our company has no AI team? 3. Which of our core systems have prebuilt connectors, and which would rely on public endpoints? 4. Will the daily query quotas cover our heaviest users? 5. How should we structure projects and licenses across several portfolio companies?

Verdict

Gemini Enterprise is a well documented choice for portfolio companies that already live on Google. Its entry price for Business is $21 a seat monthly⁷⁴, Google lists many certifications³⁹, and its data controls include region choice and customer managed keys⁶⁶⁴⁰. It ranks fifth for private equity portfolios because agent building sits with the customer¹¹, because a full deployment involves buying two products⁸, and because larger edition prices are not public⁸⁶. A fund that wants measured results inside a hold period, delivered by people who build alongside the company, will look first at Nexus.