It’s All About AI, Not Hardware (Premium)

Microsoft is hosting a special event this Thursday in which it will detail its “latest AI innovations across the company … Microsoft 365, Surface, Windows, Bing, and more.” It’s not about announcing new hardware—some continue to insist this is a “Surface event”—but is instead something far more meaningful to both Microsoft and its customers, a chance to learn how the software giant intends to infuse AI capabilities into the client products and services that you and I care about the most.

And, um, Bing. But I kid.

Yes, the presence of Microsoft 365 and Bing bring with them some sense of the cloud. But don’t get distracted by that, I don’t think this is about Azure-powered services or giant corporate entities in any meaningful way, Microsoft has its Build and Ignite conferences for those audiences. I believe that this event is about the software and services that we all use as individuals. And I am very keen to learn how Microsoft intends for us to take advantage of the AI revolution that it kickstarted.

We learned about Microsoft’s strategy here at a high level at Build 2023 in May. Stevie Batiche’s appearance at the show was a master class in communication—in sharp contrast to his former boss—but the key takeaway for purposes of this discussion was his description of how three emerging new AI application structures are shaping how AI capabilities will be delivered to users. That is, AI will be beside applications, inside applications, or outside applications.

Microsoft will deliver new AI capabilities to existing applications and services using that first structure, “beside,” via what it calls copilots. This is a perfect name, as you, the content creator, is the pilot and these UIs are helpers.

“It’s very appropriate that the first types of significant AI experiences are copilots because it enables us to get in the game quickly,” Batiche said at Build 2023. “It keeps the original app architecture definition and is minimally disruptive to what our customers already know. Yet, this new application structure delivers immensely capable tools and experiences that did not exist before.”

And with respect to the products that we know will be part of this coming event, that means the Microsoft 365 Copilot and the Windows Copilot explicitly, but also Bing Chat. Which, frankly, should be renamed to Bing Copilot.

So let’s start there. Bing’s AI capabilities are probably well-understood by this point, given the publicity. But the short version is that Bing Chat offers ChatGPT-based chat-based conversational capabilities, and can answer complex questions, and there are separate services for generating images and composing drafts of important documents, emails, and the like in a variety of tones and writing styles.

Less well known, Microsoft is integrating Bing Chat elsewhere throughout its emerging client-side ecosystem too. Bing AI capabilities first appeared in the Microsoft Edge sidebar, for example, and they are coming to the Windows Copilot as well. I suppose you consider these capabilities as foundational or core to the entire family of offerings. Anywhere at any time, you can use AI to answer questions and get help.

Windows Copilot, today, is the murkiest of the client copilots. It was announced at Build in May and promised in preview form in June, but it barely came in under the wire with its inclusion in a Windows 11 Insider Preview Dev channel build on June 29. It was, in a word, disappointing. Aside from the Bing chat and compose capabilities noted above, it offered only weak, text-based Windows integration for a handful of features like Dark mode and screenshots. Over time, it came to the other Insider channels but was never updated in any meaningful way, and yet I expect it to appear as a key new feature of Windows 11 version 23H2 in the next month or so.

Microsoft laid out an articulate vision for Microsoft 365 Copilot earlier this year, before Build, and it has since revealed that it will be augmented with a plug-in extensibility model similar to that offered by OpenAI’s ChatGPT. But despite a lengthy list of interesting capabilities—in contrast to Windows Copilot—Microsoft 365 has been in private preview so far this year, so I haven’t tested it yet. But the most exciting news here, perhaps, is tied to that plug-in model: At Build, Microsoft also said that it is partnering with OpenAI on an open standard for plugins that will work across ChatGPT, the Microsoft 365 Copilot, and the upcoming Windows Copilot.

And this, I think, is the logical place to start when speculating about this week’s AI special event. Microsoft’s goal is to spread AI across its various products and services quickly and seamlessly. And then ramp up the addition of regular new features, just as it did with Microsoft Teams, but not just for individual products and services but across the ecosystem. The open standard for plugins is key to that.

So I expect to see a host of Microsoft and third-party plug-ins/capabilities that will work across all three of these copilots, for Bing Chat, Microsoft 365, and Windows 11, respectively. And because Windows is so ill-served by Windows Copilot as it now stands, it further makes sense that this event will be the moment in which it catches up. Consider, for example, the extensibility models that exist today in Windows, where cloud storage services and third-party utilities can integrate into the File Explorer shell and provide specific new functionality (“Open with [app],” “Add to archive [with app],” and so on) and further customize the UI. What might AI-infused apps be allowed to do?

Stevie Batiche sort of addressed this at Build when he described the third AI application structure, “outside.” In this model, AI will orchestrate across multiple apps, plug-ins, and services, and will function like an agent. It will connect, orchestrate, and keep contexts across entire workflows, across devices, and even across vastly different timescales, he said.

“The Window shell is an orchestrator,” he said, “In fact, maybe one of the most powerful orchestrators across apps, across content, across the graph. Imagine with AI and natural language, you start to see glimpses of the opportunity with the Windows Copilot. Here when you get intelligence that’s functioning, not at just the granular details, but at the higher levels where you get a mixing of both tactics and strategy, you get both vision and execution. It’s like a copilot of copilots, a very powerful application structure. With the plug-in model from Bing, Windows, and Microsoft 365, we’ll start to see these outside structures emerging.”

I know, we skipped the second structure. But that was purposeful, as we will also see new AI-centric products and services that fall under the model. And I bet we see examples of this type of solution at the event too. Indeed, some already exist.

“In the second application structure, the AI is inside as the main scaffolding of the app,” Batiche said back in May. “It’s the main input loop. Here, you use AI to completely redefine the application interaction model and even its purpose. The interaction model will be less dependent on point-and-click commands. Things will become much more automatic. We see glimpses already happening in applications like Designer, Clipchamp, and Luminar Neo, that take pro-level skills and turn them into one-click, slider-driven intents, and much more intuitive interactions, all without compromising the result. Here, there are fewer toolbars, fewer deep menus, simply because you don’t need them.”

Two things come immediately to mind here.

First, his mention of Clipchamp is on point: I switched to this wonderful web app in July after Adobe’s supposedly more powerful (but also more traditional) video editing application failed me hard, and its simple, intuitive interface has kept me using it: I have created literally dozens and dozens of videos with Clipchamp since then. And if this kind of thing is the future, I’m all for it.

Second and perhaps more profoundly, I spent an inordinate amount of time wondering—OK, worrying—about the future of Windows app development, and I know some of you do as well. But it’s pretty clear that AI is literally the future of Windows app development, and that in fixating on the wrong things—frameworks, languages, native vs. cross-platform vs. web, and so on—we lost sight of what’s important. It literally doesn’t matter how developers create (or updates) apps anymore, that’s their choice. What matters is whether they’re using AI.

And on that note, consider the unexpected rolling release of AI-based features that Microsoft keeps adding to Paint, among the oldest and meekest of its Windows in-box apps. It’s clear that Paint, of all things, will factor into this special event, because Paint represents the future of all legacy apps that will be useful going forward. Indeed, it is becoming exactly what most Windows 8/10 in-box apps never were: A living breathing example of how to get it right, something that customers will actually use, and something that developers of other apps can be inspired by. (And, no, it was not always this way.) And, sure, there will be other in-box apps that get AI features too.

OK, we should discuss hardware.

No, not the three lame Surface PCs that Microsoft is expected to launch at this event for some reason. The Surface Laptop Studio 2, Surface Go 4, and Surface Laptop Go 3 are all nearly identical to the machines they replace, and two of them ship with out-of-date processor chipsets. Because Surface. Yawn.

Instead, this hardware discussion should center on AI, and on how and when Microsoft will be integrating what it called Neural Processing Units (NPUs) across its entire Surface lineup—as an insider told me was the plan earlier this year—and how that change will impact the AI experiences in Windows.

This, curiously, is something that Apple got right years ago when it started adding and enhancing the Neural Engine that’s integrated into its Apple Silicon chipsets: Many AI capabilities are significantly enhanced by on-device AI hardware, even when much of the processing occurs in the cloud. And some can of course occur fully on the device, online or not. Microsoft and some PC makers—like Lenovo, recently—have spoken only vaguely about the benefit of on-device AI hardware, but this event should be when the software giant gets more specific. Not just about the capabilities, but the timing.

And it can’t just be Surface. Microsoft’s PC family is not influential enough for the company to announce some audacious Apple-like plan to put AI into every PC it makes by some date. Instead, the company should play to the PC ecosystem’s biggest strength—choice—and explain how the entire industry is rallying around this shift, which won’t just be Windows but will include the microprocessor makers, the hardware makers, and all of the key app developers.

We’ll find out soon enough. I’ll be in New York on Thursday at will report back with my thoughts.