Prebuilt Gen AI at Work: Gemini App, Gemini Enterprise, and Workspace

A prebuilt gen AI offering is a finished product you turn on, not a platform you build with — Google has already trained the models, built the interface, wired in the safeguards, and keeps all of it updated, so the time from purchase to productivity is measured in days rather than quarters. This is the fastest and lowest-risk entry point on Google Cloud's gen AI spectrum, which runs from prebuilt products at one end, through customizable solutions, to fully custom development on the Agent Platform at the other.

The business case for starting prebuilt is straightforward: no data science team is required, costs are predictable per user, Google maintains the models and features, and the value shows up where employees already work. For most organizations, the first measurable gen AI wins — faster drafting, quicker answers, shorter meetings — come from prebuilt tools, and those wins build the confidence and skills that later justify custom investment.

The exam tests three prebuilt offerings for AI-powered work, and the reliable way to keep them straight is by scope: the Gemini app is a personal assistant with general knowledge, Gemini for Google Workspace is an assistant embedded in your email and documents, and Gemini Enterprise is an organization-wide assistant grounded in company data. Every scenario question about prebuilt offerings is asking you to match a need to one of these scopes — or to recognize the rarer case where none of them fits and a custom build is warranted (covered at the end of this lesson).

The Gemini app is Google's conversational AI assistant — available on the web and on mobile — where anyone can chat with Gemini models to brainstorm, draft, summarize, translate, analyze images, and get help with everyday questions. It is the front door to generative AI: no setup, no training data, no technical skill required. Because Gemini is natively multimodal, users can upload an image, a document, or a chart and reason about it in the same conversation as text.

Typical use cases follow the create-summarize-discover-automate pattern at an individual level: drafting an email or a project outline, summarizing a long article, preparing interview questions, explaining an unfamiliar concept, or getting a first pass at a piece of code. The business value is personal productivity with zero adoption friction — the interface is a chat box, so training costs are minimal and time to value is immediate.

For a leader, the Gemini app also serves a strategic purpose: it is the lowest-cost way to build AI literacy across a workforce. Employees who learn to prompt well in the Gemini app carry that skill directly into Gemini for Google Workspace and Gemini Enterprise, because all three share the same underlying models and conversational style.

Gemini Advanced is the exam guide's name for the premium tier of the Gemini app, which Google now sells as Google AI Pro; what the upgrade buys is capability, not Gems. Subscribers get access to Google's most capable Gemini models and a much longer context window, so a very long report, a large set of files, or a long-running conversation can be handled in one piece instead of being split up and losing track of earlier material. Higher usage limits and some newer features also arrive in the paid tier first. When an individual's problem is that the free app cannot hold or reason over enough material, the premium tier is the simple fix — no company data or new product is involved.

Gems, the capability the exam names alongside Gemini Advanced, were introduced as a premium feature but have been available to all Gemini app users, free and paid, since March 2025. A Gem is a custom version of Gemini you configure once with instructions — a role, a tone, rules, and optionally reference files — and then reuse: a brand-voice Gem rewrites any text in the company's style, a proposal-outline Gem structures every deal summary the same way, an onboarding Gem answers new-hire questions consistently. Gems do not retrain or fine-tune the model; the instructions and files are simply applied to each conversation, which is why setting one up takes minutes.

Gems have also moved beyond the standalone app. Since 2025 they can be shared with colleagues, and they can be used from the Gemini side panel in Gmail, Docs, Sheets, Slides, and Drive, so a team can apply the same specialist where it works. The business value is consistency and saved setup time: one skilled prompt writer captures expertise once and the whole team reuses it.

What a Gem is not is an organization-governed assistant. Its knowledge is its instructions plus whatever files its creator attached — a snapshot that does not stay current with live company systems — and it is owned by an individual rather than administered by IT. When an assistant has to draw on live company systems and be administered centrally rather than by one person, the step up is a Gemini Enterprise agent, not a more widely shared Gem. Sharing is not the dividing line; grounding in connected data and central governance are.

Gemini Enterprise (formerly Google Agentspace) brings generative AI to the whole organization by grounding Gemini in the company's own data — its documents, sites, and business systems — under enterprise-grade security and governance. Where the Gemini app knows the world, Gemini Enterprise knows your business. Employees ask questions in one place and get answers assembled from the organization's actual knowledge, with references to the sources used.

Its reach comes from connectors. Gemini Enterprise connects to Google sources such as Google Drive and Gmail and also to third-party repositories and applications such as Microsoft SharePoint and Outlook, Confluence, Jira, ServiceNow, and Salesforce, so one question can draw on content wherever it lives. That is the practical difference from Gemini for Google Workspace, which centres on the user's own Workspace content.

Grounding works by retrieval, not retraining. When someone asks a question, Gemini Enterprise searches the connected sources at that moment, retrieves the relevant passages, and has the model answer from them. Company files are not used to fine-tune or retrain the model, so answers reflect the current state of the sources, a newly updated policy is picked up without any training run, and nothing an employee could not otherwise see is baked into the model.

Permissions follow the asking user. For connectors that carry access controls — such as Google Drive, SharePoint, or Jira — retrieval honours each source system's permissions for the person asking, either by syncing its access lists or by querying the source with the user's own sign-in. A search, an assistant answer, or a shared agent draws only on documents that user can already open; it does not run with the access of the agent's creator or of the connector's service account. Two employees can ask the same agent the same question and get different answers because they can see different files — and an employee without access to restricted documents gets an answer built from what they can open, with nothing leaked from the rest. The exception is a data store with no access information attached, such as indexed public website content or files imported without access metadata: everything in it is visible to every user of the app, so restricted material should come in through a permission-aware connector.

Three capabilities define Gemini Enterprise on the exam, and each answers a different kind of need.

Multimodal search lets employees search across all connected company content in natural language. Because it is multimodal it spans text documents, presentations, images, and other media, so people can find content by describing what it shows or says rather than knowing its name or where it is stored. Its scope is deliberately broad: everything connected that the user is allowed to see.

Custom agent capabilities let teams create purpose-built agents — an HR policy agent, an account-brief agent, a contract-review agent — grounded in the company sources they choose. Agent Designer is the no-code builder: a business user describes the agent's job, picks its data sources and tools, and shares it with a team, with no application code and no developers. Google also provides prebuilt agents (for example for deep research), and agents built elsewhere can be made available in the same place. Shared agents stay under the organization's administration, and, as above, each user's answers come only from sources that user can open — sharing the agent does not share its data.

Gemini Notebook Enterprise (formerly NotebookLM Enterprise) is the notebook capability inside Gemini Enterprise. A notebook holds a curated set of sources the user chooses — say, a set of approved contract templates — and produces summaries, briefings, answers with citations, and audio overviews grounded only in those sources. That curated scope is the contrast with multimodal search: search looks across everything connected, superseded versions included; a notebook answers from exactly the set you chose.

The Gemini Notebook API is how a team's own internal application drives notebooks programmatically. It is a management API: it can create, get, list, delete, and share notebooks, add and manage their sources, and generate audio overviews. The API does not replace Gemini Enterprise's search or agents; it automates the curated-notebook workflow from inside software the organization already runs.

The business value is unlocking institutional knowledge. Most organizations answer the same internal questions thousands of times a year through interruptions, tickets, and tribal knowledge. Gemini Enterprise converts that buried knowledge into instant, permission-aware answers and reusable agents, which shortens onboarding, reduces internal support load, and keeps decisions grounded in the organization's own sources.

Gemini for Google Workspace embeds Gemini directly into the productivity tools employees already use — Gmail, Docs, Sheets, Slides, Drive, Meet, and Chat — so AI assistance appears in the flow of work instead of in a separate destination. In Gmail, Help me write drafts and refines messages and replies, and Gemini summarizes long threads; in Docs, Help me write drafts new text and rewrites, shortens, or changes the tone of existing text; in Sheets it helps organize and analyze data; in Slides it drafts slides and generates original images without leaving the deck; in Drive it summarizes files and folders; in Meet it takes notes and captures action items. The Gemini side panel brings Gemini — and your Gems — to the file or thread in front of you, and can reference your other Workspace content.

The defining advantage over the standalone Gemini app is context without copy-paste: the assistant works on the email thread, document, or spreadsheet the user already has open. For an individual's task on their own email or files, Workspace is the direct fit even in a company that also licenses Gemini Enterprise.

Data protection is built in and on by default. Google states that customer content used with Gemini in Workspace — prompts, files, and responses — is not used to train or improve models outside the organization without permission and is not reviewed by humans for that purpose. Workspace's existing security, compliance, and data-region controls carry over, and none of this requires client-side encryption or a Gemini Enterprise licence. Gemini also inherits Workspace access controls: it can only use content the signed-in user can already open, so it cannot summarize or quote a file the user has not been granted access to. Administrators turn Gemini on or off by organizational unit or group; that controls who can use Gemini, not which files a user can reach — file access is still governed by sharing permissions.

The business value is measured in reclaimed time across the entire workforce: minutes saved on every email, first drafts in seconds, meetings that end with notes already written. Because the gains attach to activities every employee performs daily, small per-task savings compound into significant organizational capacity.

The three offerings are complements, not competitors — they differ by where the assistance lives, what data grounds it, and who governs it, and the table below is the comparison the exam expects you to know.

Two dimensions resolve nearly every scenario question. Scope: is the beneficiary one person or the whole organization? Grounding: does the answer need general knowledge, the user's own Workspace files, or company-wide data across systems? An individual working with general knowledge or uploaded material is the Gemini app; an individual working in their email and documents is Gemini for Google Workspace; an organization searching its collective knowledge or deploying shared, governed agents is Gemini Enterprise. A frequent near-twin is the personal Gem versus the Gemini Enterprise agent: both are reusable assistants and both can be shared, but only the agent is grounded in live connected company data and administered centrally.

Prebuilt offerings are the right choice for common employee productivity — writing, summarizing, finding internal knowledge, team assistants — because they deliver value fast, need no ML team, and are maintained by Google. A custom build on the Agent Platform belongs where the AI is the product or the differentiator: a customer-facing experience inside the company's own app or website, logic that must call proprietary systems and APIs (an order system, a pricing service), or behaviour no packaged product offers.

The two are not either-or. A typical plan uses Gemini for Google Workspace or Gemini Enterprise for the workforce and invests its scarce engineering capacity in one differentiated customer-facing build on the Agent Platform. Building a custom tool to replicate what a prebuilt product already does — an internal drafting assistant, an intranet search bot — spends that capacity on something the organization could simply turn on. (Developer tooling and customer-experience offerings are covered in their own lessons in this domain.)

A mid-sized consulting firm shows how business value grows with scope — personal productivity, then workforce time savings, then institutional knowledge — and how the three offerings enable AI-powered work together. The firm starts with Gemini for Google Workspace because it already runs on Gmail and Docs: consultants draft client emails faster, summarize long project threads before status calls, and let Meet capture action items — the quickest measurable win, since it attaches to work everyone does daily, with the firm's existing Workspace protections applying from day one.

Individual analysts use the Gemini app to pressure-test ideas and explain unfamiliar topics, and a partner builds a Gem that lays out every proposal in the firm's standard structure and tone and shares it with the team, who use it from the Docs side panel; proposal quality becomes consistent without a style-guide policing effort. Analysts who work with very long industry reports move to the premium Gemini app tier for its larger context.

Finally, the firm deploys Gemini Enterprise to solve its oldest problem: fifteen years of engagement reports, methodologies, and templates scattered across Drive, SharePoint, and Confluence that new hires cannot find. Multimodal search lets any consultant ask what the firm has done for clients in a given industry and get grounded answers with sources; the knowledge team uses Agent Designer to build a methodology agent that answers how-we-do-it questions, and each consultant sees only engagements they are cleared for. The firm's project tool calls the Gemini Notebook API to create a notebook from each finished engagement's approved deliverables and share it with the next team. Onboarding time drops, senior staff field fewer interruptions, and institutional knowledge outlives employee turnover — three offerings, three scopes, one adoption path.