As I watched the sessions at this week's Google Cloud Next, what interested me most were the demos of how the company's new Duet AI feature is being incorporated in its core Workspace productivity suite. I'm also impressed by how generative AI improves other features such as development and cloud security. But since I spend much of my time in office applications, this grabbed most of my attention, as it did when Microsoft demoed similar "Copilots" for Office 365 earlier this year.
At this year's conference, Google said that thousands of companies and over one million people have used beta versions of Duet AI for things such as generating emails about late invoices and writing blogs. It said that Duet AI is now generally available for Workspace at $30 per month per user. It's surely not a coincidence that it is the same pricing that Microsoft earlier announced for Copilot for Office 365, although Microsoft's solution is not yet generally available. I expect it will be shortly. (Google Workspace still has far fewer paying customers than Microsoft Office. At the conference, Google said Workspace has 10 million, while Microsoft claims that Teams has more than 300 million monthly users, suggesting that Office 365 has considerably more.)
Google's Duet AI demos included producing a creative brief for a marketing department, in which it suggested templates and generated a presentation based on the company's corporate Docs and Sheets.
It showed a new teleprompter feature within Google Meet, with Duet AI taking notes during a meeting and creating summaries about it. The biggest change showed it moving away from prompts and to contextual interaction, such as suggesting creative ideas, or even taking actions on your behalf.
The demos themselves were impressive, as were Microsoft's earlier ones. I'm looking forward to trying them both out.
But I have several big questions, the same I expect most corporate IT executives will have before signing up. The biggest is whether employees will get $30 a month of value from these generative AI features. I do not doubt that some will—it's easy to see how this can help in creating marketing presentations or drafting sales emails. But for others, I'm not so sure.
On one hand, $30 per user per month sounds expensive contrasted with the price for the basic Workspace or Microsoft 365. Workspace pricing starts at $6 per user per month, though with very limited storage, going up to $18 per user per month with better management and more space. Microsoft 365 (the new name for Office 365) also starts at $6 per user per month for basic functions, up to $22 per user per month, including desktop versions and advanced security. Of course, either company will negotiate larger enterprise agreements. But it certainly sounds like adding the AI features more than double the pricing. AI is nice, but is it that much more valuable than email, word processing, spreadsheets, and presentations combined?
On the other hand, we're told that training large language models and then running AI conversations against the applications based on said LLMs is expensive. For many office workers, $360 a year for something that can summarize meetings, suggest text, or convert spreadsheets into slides seems very worthwhile. After all, that's a lot less than you'd pay for a human assistant.
Many businesses will see this as worthwhile for some users, but not others, and that may complicate licensing. Right now, most organizations I know tend to have just one or two licenses that cover their whole workforce—possibly a full license for office workers and a more limited one, focused on email, for front-line workers.
It's not clear to me that all office workers will be able to take advantage of the generative AI features. I see where these features will help marketing and sales professionals in drafting emails or creating presentations. But will it help people who spend most of their day doing engineering, processing invoices, filling out tax forms, etc.? I do expect that generative AI features will be part of new applications or added to existing specialized software, and it may well be very useful there; for instance, one of the big early users seems to be software developers. That doesn't mean the people who spend their days doing those tasks will use such features in Workspace or Office.
One of the first broad uses of the technology will likely be in customer service applications and chatbots. This could have a huge impact on those jobs, while hopefully improving customer service. I'm not sure that the use of typical office productivity software will matter as much for these people. As for jobs in factories, or technicians, or health care professionals, the benefit of AI mostly resides in the applications tailored to their work.
Finally, there remains a lot of concern among the CIOs I talk to about protecting organizations' intellectual property. Both Google and Microsoft say they are not going to use corporate data to train future iterations of their AI products, but organizations will want to look over the agreements closely. That doesn't begin to deal with the questions surrounding how the underlying models were trained in the first place. Of course, smaller organizations and individuals will have to accept the license as written if they want to use the new features.
Again, I want to see the final versions of the software before making a judgment, and I expect many CIOs and IT executives will want to try this with some users before making a decision. Still, generative AI—whether in Duet AI or Copilot—represents the most consequential change to productivity software in many years. I'm waiting to see how we all make use of it.