How to Use Google AI: Gemini, Workspace, and AI Studio

How to use Google AI comes down to picking one of three doors: the Gemini app for asking and drafting, Gemini inside Google Workspace for work that already lives in your Gmail and Docs, and AI Studio plus the Gemini API for anything you want to build. Most people only ever open the first door, which is why they conclude Google AI is another chatbot.

TL;DR

• Three doors, three jobs. App for chat, Workspace for work you already have, AI Studio for anything you ship.
• The scale is real. On Alphabet's Q2 2026 earnings call (July 22, 2026), Sundar Pichai reported 950 million monthly active users on the Gemini app, daily actives tripled year over year, and more than 9 million developers building monthly with Google's models.
• You are probably paying for it twice. Gemini is bundled into Google Workspace from the $7 per user Starter plan up.
• Flash is the default. Gemini 3.8 Flash is $0.75 in and $3.75 out per million tokens through Dec 31, 2026, then doubles. Flash-Lite is $0.30 and $2.50.
• Free has a price. Google's pricing page states free-tier content is used to improve its products. Paid-tier content is not.

The three doors into Google AI

Google does not sell one product called Google AI. It sells the same family of Gemini models through several surfaces, and the surface you pick decides your cost, your data terms, and what the model can actually see.

That last part matters more than model quality. A brilliant model with no access to your inbox loses to a decent one that can read the thread you are replying to.

Door one: the Gemini app, and what a subscription buys

The Gemini app is free with a Google account and capped on access to the top model. The paid tiers do not unlock secret intelligence. They raise limits and bundle other things.

Google's AI plans page sets it out by multiplier. AI Plus gives roughly 2x the free access to Gemini 3.1 Pro plus 400 GB of storage. AI Pro gives 4x access, 5 TB of storage, Deep Research, the Gemini Spark personal agent, and $10 a month in Google Cloud credits. AI Ultra, which Google introduced at I/O on May 19, 2026 at $100 a month, goes to as much as 20x access, 20 TB of storage, higher access to Deep Think reasoning, and $40 a month in Cloud credits.

Read that as a usage ladder, not a capability ladder. If you hit limits twice a week, upgrade. If you hit them twice a year, do not. The same logic applies to every assistant subscription, which is the argument I make in how to choose an AI model: match the tier to the job, then stop paying attention to the marketing.

Door two: Gemini in Workspace, which you already bought

This is the door people skip, and it is the one with the best return.

Gemini is bundled into every Google Workspace business plan. At $7 per user per month on Starter you get the Gemini assistant in Gmail and basic app access. At $14 on Standard it reaches Gmail, Docs, Meet and more, with expanded access to Google's most capable models. At $22 on Plus you get full model availability in the Gemini app plus expanded Gemini Notebook. Those are the annual-commitment prices from Google's own Workspace pricing page.

The features are concrete rather than vague. Draft emails and summarize threads in Gmail. Help me write in Docs. Create tables and AI-enhanced Smart Fill in Sheets. Help me design in Slides. Take notes for me in Meet, which writes the meeting summary without a bot joining the call. Side-panel access in Drive that reads the file you have open.

Here is the practical difference. In the Gemini app you paste a document in. In Workspace the model already has the document, the thread, the calendar entry, and the file permissions attached to your account. That removes the copy-paste step that quietly kills most AI habits by week three.

If you run a team on Workspace and have never audited who has Gemini switched on, that is the highest-value hour on your calendar this week. Against the backdrop of the US Census Bureau's Business Trends and Outlook Survey, which put national business AI use at 19.8% as of May 3, 2026 (39.7% in Information, 33.9% in Finance and Insurance, 37% among firms with 250 or more employees), flipping a switch you already pay for is a cheap way to move.

Door three: AI Studio, where free costs you something else

AI Studio is where you stop chatting and start building. You write a prompt, run it against a specific model, watch the token count, then generate an API key and take it to code.

Two things to know before you put anything real through it.

First, the data terms. Google's Gemini API pricing page says free-tier content is used to improve Google's products, and states for every paid tier that content is not used to improve Google's products. That is the actual difference between free and paid, and it matters more than the money. Attaching a billing account moves you from free to Tier 1 more or less instantly according to Google's rate limits documentation, so there is no excuse for running client data on a free key.

Second, the tier ladder is spend-gated, not application-gated. That same page states Tier 2 requires $100 or more in cumulative Google Cloud spend plus three days from your first payment, and Tier 3 requires $1,000 or more plus 30 days. Plan for that if you expect a launch spike, because you cannot buy your way past the waiting period on the day you need it.

Picking a Gemini model without reading the whole docs page

Google's models page currently lists dozens of entries across text, image, video, audio, music, embeddings, and robotics. You need three of them.

Run the arithmetic before you pick. Say you classify 50,000 support tickets at roughly 800 input tokens and 150 output tokens each. That is 40 million input tokens and 7.5 million output tokens.

• Flash-Lite: 40 x $0.30 plus 7.5 x $2.50 equals $30.75
• Gemini 3.8 Flash: 40 x $0.75 plus 7.5 x $3.75 equals $58.13
• Gemini 2.5 Pro: 40 x $1.25 plus 7.5 x $10.00 equals $125.00

Four times the price between the cheapest and the most expensive, on a job where the output is one of six labels. That gap is the entire argument for building a small eval set first. The full version of that argument is in how to test an AI agent, and the cost side is in cutting your AI API bill.

One more line item people miss: the Flash pricing above is promotional through December 31, 2026, after which Google's pricing page lists $1.50 input and $7.50 output. Same model, double the bill, on a date already published. Budget for it now.

Where Google is not worth switching to

Three honest limits.

The 1M context window is not a retrieval system. Google's own long context documentation reports roughly 99% accuracy on single-needle retrieval, then says plainly that when you are looking for multiple specific pieces of information the model does not perform with the same accuracy, and that longer queries mean higher latency. Stuffing a million tokens in is a convenience, not an architecture. If you need reliable recall over a corpus, you still want retrieval, which is what chunking documents for RAG is for.

Model IDs move. That same models page lists Gemini 2.0 Flash, Gemini 2.0 Flash-Lite, the Gemini 3.1 Flash-Lite preview, and the Gemini 3 Pro preview as shut down. Pin exact model IDs in production and read the changelog before you upgrade. This is true of every provider, and it bites hardest on the one shipping fastest.

Switching is an eval project, not an API swap. If your prompts, tools, and tests are already tuned on another provider, the cost of moving is measured in re-validation, not in rewriting a client. I said the same thing about the other direction in the OpenAI API guide. The right move is to run both against your own eval set on your own tasks, then decide with numbers instead of vibes.

The bottom line

Using Google AI well is a routing decision, not a subscription decision. Ask and draft in the Gemini app. Do work that touches your real documents in Workspace, where you are almost certainly already paying for it. Prototype in AI Studio, then ship on a paid API key with the model ID pinned and an eval set behind it. Start at Flash, verify with your own tests, and only pay for Pro on the tasks that measurably need it.

Want the rest of the stack? The AI Daily Driver Stack is the exact set of tools I run every day, with what each one is actually for. And if you want this kind of breakdown before it goes stale, join the newsletter. One issue a week, no fluff.