How Google is integrating Gemini into Workspace

In January, Google made Gemini — previously an add-on — a standard feature for business users in Google Workspace. That move made a point: generative AI (genAI) isn’t an add-on, it’s a fundamental capability that should fit naturally in the flow of work. 

GenAI is now more visible (and doing a lot more) in Workspace, including suggesting document drafts, automating workflows, analyzing documents and even adding videos and avatars to presentations.  Many new features were added at the recent Google IO and Cloud Next shows.

Google’s cloud-first Workspace AI differs from Microsoft’s approach of building a family of models and agents that excel at specific tasks or workloads.

srcset='https://b2b-contenthub.com/wp-content/uploads/2025/06/Kristina_headshot-1.jpeg?quality=50&strip=all 512w, https://b2b-contenthub.com/wp-content/uploads/2025/06/Kristina_headshot-1.jpeg?resize=150%2C150&quality=50&strip=all 150w, https://b2b-contenthub.com/wp-content/uploads/2025/06/Kristina_headshot-1.jpeg?resize=300%2C300&quality=50&strip=all 300w, https://b2b-contenthub.com/wp-content/uploads/2025/06/Kristina_headshot-1.jpeg?resize=168%2C168&quality=50&strip=all 168w, https://b2b-contenthub.com/wp-content/uploads/2025/06/Kristina_headshot-1.jpeg?resize=84%2C84&quality=50&strip=all 84w, https://b2b-contenthub.com/wp-content/uploads/2025/06/Kristina_headshot-1.jpeg?resize=480%2C480&quality=50&strip=all 480w, https://b2b-contenthub.com/wp-content/uploads/2025/06/Kristina_headshot-1.jpeg?resize=360%2C360&quality=50&strip=all 360w, https://b2b-contenthub.com/wp-content/uploads/2025/06/Kristina_headshot-1.jpeg?resize=250%2C250&quality=50&strip=all 250w' width='512' height='512' sizes='(max-width: 512px) 100vw, 512px'>Kristina Behr, Google’s vice president of product management for Google Workspace.
Google

Computerworld sat down with Google’s Kristina Behr, vice president of product management for Google Workspace, to explore where Workspace is headed with AI.

How are you making it easier for users to access AI in Workspace?  “A big UX paradigm that we’re driving is — how do we bring Gemini thoughtfully into the places that users are already working? So, no switching apps, no disrupting their rhythm, just practical help in the moments that matter.

“I’ll go back to some OG examples. Smart Reply and Smart Compose or malware and phishing protections in Gmail are pre-LLM examples of things that were just helpful. People didn’t care if it was AI or not; they just wanted to respond to their email faster.

“It’s that same thinking that makes Google AI and Gemini inside of Workspace so accessible. We want to get to this place where we understand what your pain points are, sort of like the drudgery in your work, and it’s just going to help you knock it down.

“It may not look like a personified AI agent. It may look like an easy tab forward in order to get that suggested reply. It’s like killing two birds with one stone, really helping users decrease the fear and hesitancy around AI by just making it so useful that people will use it every day.”

Right now, Workspace entry points are Docs or Sheets or other apps. Will AI change the interface or will integration be within the apps as we know them? “Users can access AI in Workspace apps in two different ways. One is a more prompt-driven experience in the side panel available across about seven different products in Workspace. It’s a little bit more user-driven. The prompt box is very powerful in that you could add specific files to ground the response.

“The AI that we’re investing in…the interface will be accessible in the main canvas. We want to be in the flow of work. 

“Imagine you’re writing a document, and then you ask Gemini to suggest some refinements. It suggests you can make something punchier, and help you like a buddy would. We’re taking advantage of current user experience paradigms, but it’s an AI instead of a friend.  Some of it will be auto-suggested, like the Gmail experience, and some of it will maintain that side panel experience.”

Do you see the lines blurring between Sheets and Docs as data comes together? At some point, can users just say create a document as opposed to saying create a Doc or Sheet? “You’re bringing up an awesome point. Workspace is a suite, not a point solution. We’re a platform creating tools that work really well together.

“Docs, Sheets, and Slides are the primary ones. We’ve been making collaboration easier, and now it’s better because of AI. AI doesn’t change this. It amplifies our thesis that tools are stronger together.

“One of the examples is…take notes for me in Meet. Those are based off of a Doc. Instead of having Meet go create its own thing and then have meeting notes in a siloed experience, it’s just using Docs. How do we make that accessible in Drive and automatically linked to your calendar? Those synergies showcase our secure cloud-first strategy.

“Will all these converge into some uber app? I don’t think so, near term. There’s utility in a fantastic spreadsheet tool. You can draft emails based off Docs. Write Docs based off Slides. Those are the types of things we’re enabling.”

Google is cloud-first, but is there any value for you to have offline AI models for Workspace users? “We are a cloud-first product. Users collaborate with Gemini when they are connected to the Internet. Users can use our tools without an Internet connection — you can view, edit, create.

“We are constantly re-evaluating and thinking about how we make this powerful technology genuinely useful. But for now, Gemini users need to be connected to the internet.”

Right now, it’s Gemini. Are you looking to make smaller models available, like Gemma 3N, for customers who might need it in Workspace but don’t want Gemini or prefer a custom model that can be built off their own data? ”To serve users across the globe, we want to simplify this as much as possible. We think in terms of what are folks trying to do? What’s the critical user journey? What’s the job to be done? How do we match the model for that use case?

“So instead of like a manual transmission car where our customers need to pick the model based on what we’re trying to do, let’s just skip that step for them. And if what they’re trying to do is analyze a spreadsheet, what’s the best thing for us to do so that we can focus on helping them analyze the data, surfacing the key insights, and not worry about specific models?

“This is particularly true considering the breadth of our customer base. That very specific model selection capability is generally a thing that Vertex and our Google Cloud folks focus on, which is giving their users more control since their user base is different.”

Google recently introduced “AI Mode” for Search. It doesn’t seem related to Google Workspace, but is there an idea of how Google will merge or integrate it?  “At a fundamental layer, we’re all based off the same cool models and tech that Google DeepMind is creating. What every product in Workspace is thinking about is how they take those best-in-class large language models and make really great experiences for those particular products.

“Is there going to be, some day, a potential grand unification? I don’t know — maybe. We’re focused on making Gemini in Workspace genuinely useful. There will oftentimes be alignment at the core tech level, because we’re all based off the same capabilities. There will be natural alignment, even if it’s not the exact same product.”

I feel I need to use my data in Google Drive to make Gemini in Workspace more effective. How can people take advantage of their own data in Google Workspace? “You’re hitting on one of the best value props for Workspace. It is a suite. By having Drive and Gmail and Docs together, you can imagine powerful user experiences in a privacy-compliant way.

“So, you’re drafting that response in Gmail, the fact that it knows all the dates on the project that you’re working on from a file stored in Drive, that is super powerful.”

Gems is really interesting. Could you talk about its functionality? “Gems is just very plainly a way to customize or personalize Gemini’s responses. We have folks creating Gems that are interview prep guides or various different things.

“Right now, you create a Gem through tips using natural language. We have lots of tips and tricks to help people create these Gems.

“We have users that love it. Examples that come out of the box include a brainstorming partner, coding partner. Then people can write their own.

“We see it most popular in large organizations where they want to help employees get more standard capabilities out of Gemini by giving it more instruction.”

I tried Gems — it’s menu and text-driven. Can I just tell it what I want and it’ll figure it out automatically? ”It’s text-based. But you can just use natural language. We’re not expecting people to be developers or need to know how to code to create it.

“When you write your instructions, you do have to type it. The audio input is a great idea. We can keep you updated as different input things come in. But just normal natural language, no command prompt needed.”

With Gems, I also felt I needed to know more about company processes and procedures. “We thought about workflows by going back to what’s time consuming in people’s day. How can we be really useful?

“There are some traditional automation tools that handle simple triggers, like if-then logic. One of the things we’ve been noodling on is what if work requires context and reasoning? That’s where Google Workspace Flows comes in. It’s a new way to automate those multi-step processes that can use AI to actually research, analyze, and generate content.

“What’s powerful is today we have customers getting feedback from their own customers, then they need to respond and coordinate with colleagues. We start with a real customer problem and span out from there versus taking a cool new piece of tech, shipping it, then backing into utility.”

How do you ensure data security? “Data security, confidentiality, compliance are top priorities for business leaders adopting AI. We’re committed to keeping your data safe. Your data is your data. We don’t use your data, prompts, or responses to train Gemini models outside your domain without permission. We don’t sell data or use it for ad targeting.

“You’re always in control. We built Google Workspace with Gemini with enterprise-grade controls to roll out Gemini while protecting sensitive data. Gemini only retrieves data the user has access to. Your existing Workspace security controls are automatically applied.

“You stay compliant. Gemini is the first generative AI solution to attain comprehensive safety, privacy, and security certification.”