Gemini for Google Workspace and the Gemini app

The second session started with a complaint. Dana Leclair, Birchline's chief financial officer, had forwarded Farah a thread from the store operations team: staff were pasting supplier price lists into a free consumer chatbot to have them summarized. "I don't know what happens to that data," Dana wrote, "and I don't want to find out from a newspaper." Farah read it aloud to Owen and Jules. "This is why we start at the top of the stack," she said. "People already want these tools. If we don't give them a safe, approved version, they'll use an unsafe one."

This lesson covers the two most widely used Gemini products for individuals at work: the Gemini app and Gemini for Google Workspace. Both sit at the application layer. Neither needs a developer. The decisions a leader makes about them are about licences, settings, data protection, training and measuring value.

The Gemini app

The Gemini app is Google's general-purpose AI assistant, available on the web and on phones. A person types or speaks a request and Gemini responds: it can answer questions, draft and rewrite text, summarize long documents, brainstorm, explain code, analyse an uploaded spreadsheet, describe an image or help plan a project. Some features, such as deeper research that reads many web sources before answering, are available in higher tiers. The exam guide refers to the app together with its premium tier, which Google has sold under names such as Gemini Advanced; as with every product name in this course, check the current plan names before buying.

For an individual consumer, the Gemini app is personal. For a company, what matters is the business edition: when employees sign in with their work accounts under a company subscription, Google applies enterprise data protections. Google's stated commitments for business editions include that company prompts and responses are not used to train Google's models outside the company's domain and are not reviewed by people for that purpose, and that administrators control the service. Farah's point to Dana was simple: the same kind of tool the store team was already using, but under Birchline's accounts and Birchline's rules.

Gems: saved instructions for repeated tasks

A Gem is a customized version of Gemini that a person creates with saved instructions, and sometimes with reference files, for a task they do often. Instead of typing the same long prompt every week, the person opens the Gem and gives only the new material. Gems are a lightweight way to package good prompting for colleagues who will never read a prompt engineering guide.

Jules built one during the session. His Gem, called Product copy reviewer, had instructions to check a draft product description for Birchline's house style: plain language, metric and imperial sizes together, no claims about waterproofing unless the product had passed the lab test, and a limit of 120 words. Its reference file was the style guide. He pasted in a draft for a new rain shell, and the Gem flagged the phrase "totally waterproof" and suggested "waterproof to the rating on the label".

Gemini for Google Workspace

Gemini for Google Workspace puts Gemini inside the tools people already use. For Birchline, which runs on Google Workspace, this was the most obvious place to start, because staff would not need to learn a new app or copy material between windows. Google has changed how Gemini is packaged with Workspace more than once, from separate add-ons to features included in Workspace business and enterprise plans, so a leader should check the current plan before budgeting.

A side panel in each app lets a person ask Gemini questions about the open file or about other files they can access in Drive and Gmail. Gemini respects Workspace sharing: it can only draw on content the person is already allowed to see. Farah stressed this to Teresa, who had worried that Gemini might show a store manager the payroll file. "It can only show them what they could already open themselves," Farah said. "Which means our sharing settings matter more than ever. If payroll is shared with everyone by mistake, Gemini will find it faster than a person would."

Data protection in business editions

Leaders deciding whether to approve Gemini at work usually ask the same five questions. The answers below summarize Google's published approach for business and enterprise editions in general terms; the contract and the current data protection documentation are what a company actually relies on.

Teresa added one more check of her own. Birchline already had data loss prevention rules in Workspace that blocked emails containing full payment card numbers. She confirmed that the same classification and data protection settings still applied when content was created or summarized with Gemini's help, and she wrote a one-page guide for staff: what to use Gemini for, what not to paste into any tool, and how to report a strange answer.

Measuring value

Dana wanted a number before paying for licences across 1,100 staff. Farah proposed a pilot with 60 people from head office and four stores for eight weeks, with three measures: time saved on common tasks, measured by short weekly surveys and a few timed tasks; adoption, measured by how many pilot users used Gemini at least weekly, which Workspace reports show to administrators; and quality, measured by a sample review of drafts by managers. The pilot runs from January to early March 2027, so its results are still to come; the model below uses illustrative figures to show the arithmetic.

# Illustrative pilot arithmetic (assumed figures, not results).
pilot_users = 60
minutes_saved_per_week = 45        # assumed average from surveys
weeks_per_year = 46                # working weeks
loaded_cost_per_hour = 48.0        # illustrative cost of an hour of staff time
seat_cost_per_user_year = 240.0    # illustrative cost per user

hours_saved = pilot_users * minutes_saved_per_week / 60 * weeks_per_year
value = hours_saved * loaded_cost_per_hour
cost = pilot_users * seat_cost_per_user_year
print(f"Hours saved per year : {hours_saved:8.0f}")
print(f"Value of that time   : {value:8.0f}")
print(f"Cost per year        : {cost:8.0f}")
print(f"Value per unit of cost: {value / cost:6.1f}")

Hours saved per year :     2070
Value of that time   :    99360
Cost per year        :    14400
Value per unit of cost:    6.9

Farah warned against reading too much into one ratio. Time saved is only valuable if it goes somewhere useful, such as more time on the shop floor, and surveys tend to overstate savings. She also wanted the pilot to record where Gemini was not helpful, because that would shape training. "A tool that saves forty minutes for half the people and wastes ten for the other half needs better training, not a bigger rollout," she said.

Rolling out to people, not just accounts

Turning on a feature is easy; changing how people work is not. Farah's rollout plan had five steps, which also match how the exam thinks about adoption.

• Prepare the ground. Clean up file sharing, confirm data protection settings and write a short acceptable-use guide.
• Start with a pilot group. Choose volunteers from several teams, including sceptics, so the results reflect real work.
• Train on real tasks. Short sessions built around each team's own tasks, such as store schedules or supplier email, teach more than a general demo.
• Name champions. One person in each team collects tips, shares good prompts and Gems, and passes problems back to IT.
• Measure and adjust. Review usage reports and survey results before expanding, and drop or change what is not working.

Where the Gemini app ends and other products begin

Jules asked whether the Gemini app could become the shopping assistant. Farah said no, and explained why the exam often asks a version of this question. The Gemini app and Gemini in Workspace are for employees working with their own files and email. They are not designed to be placed on a public website, connected to Birchline's order system or given rules about which products may be recommended. Those needs belong lower in the stack, on the AI platform. Similarly, if staff want to search across many company systems at once, not just Drive and Gmail, Gemini Enterprise is the better fit, which is the subject of lesson 3.

Dana approved the pilot that afternoon. Teresa started the sharing clean-up, and found eleven folders shared with everyone in the company that should not have been, including an old store floor plan with alarm codes written on it. "That one would have been bad with or without AI," she said, and fixed it.