I get asked this a lot: “Is SaaS dead?” — and, given that all of the PLMish companies in our space sell their software that way, it’s a good question.

The whole “SaaS is dead” thing apparently started when Microsoft’s CEO, Satya Nadella, said something like this: traditional Software-as-a-Service (SaaS) models will be replaced by AI-driven, agentic, highly integrated platforms (like Microsoft’s, obviously). Since these platforms won’t have applications as we understand them today, the modes of selling them as we know them today would be gone.

Notice that he’s not saying the apps will be gone; he’s saying how we interact with them will change. Instead of clicking on an app to launch it, an AI agent would orchestrate the task we want to do in the app. So, rather than opening MyCAD, we’d say something like “notify Supplier A that this flange has to be 10cm longer” and an agent would know that means changing a CAD model, recreating a drawing (or the PMI or whatever) and issuing it to the supplier with a change notice, request for quote and schedule and so on, whatever else must happen to make that design change into reality.

CAD is still an integral part of the agent’s process. And how do we ensure that the CAD app is accessible? By buying a license. Likely under a SaaS subscription. (And even though Im using CAD as an example here, the same logic applies to CAE, CAM, PDM, PLM, etc.)

In this scenario, the human doesn’t open or use the CAD product and, therefore, doesn’t engage with the MyCAD user interface (UI) at any time, only the AI agent’s UI. But say we’re using an AI agent in MyCAD to define and execute the task. The traditional UI “mouse-click here, then here, then enter a number here and hit Enter” interface will likely change, since an agent embedded in the CAD product will handle much of that work for us. The old click-click-click that we’re all used to is hugely inefficient, even for the expert users who are rapidly leaving the workforce. For newbies, this UI is often too complicated, even when palettes and other paradigms simplify it. Adding an AI agent and its UI to the MyCAD UI will reshape all of our user interfaces — but it’s still driving the MyCAD software, likely purchased via SaaS.

That leads to the next question: so can’t I use AI to create an app specific to my needs? Get out from under the bill for commercial SaaS tools? Perhaps — but are you sure this is a good idea? Yes, software developers are using agents to write code; that capability will make its way to non-expert user communities, too. But commercial software vendors are in the business of creating commercial-grade solutions, with professional coders, support staff, training partners, and so on. They create at scale, for hundreds or thousands of users, and can amortize costs over a wide base. I can see myself using AI to create a UI for MyCAD that only shows me the parts of the app I use all the time (and for which I can’t use an AI agent right now) — but I can’t see that extending to all of the complicated bits and pieces that make CAD software, CAD.

Remember too that API calls to AI engines are still very, very expensive. I’ve seen $400,000 bandied about as a possible price tag for creating a notional app that’s far simpler than a fully functional CAD. Is having your own personal tool worth that much? And what happens when something needs to be added or changed in the app? Will that cost another $400,000? Yes, the costs of AI API calls will come down, but do you and I really want to get into this business?  If my job is making snow plows (it’s snowing here as I write this), I’m better off doing that and leaving CAD tool development to the experts. 

(That’s not to say that new commercial CAD vendors won’t appear, coding quickly with AI to build a rich feature set and using all sorts of modern API technologies to hook into existing ecosystems. I wouldn’t be surprised if we see a number of those. But I don’t think DIY-CAD will take off. It’s just too complicated. Of course, LMK if I’m wrong; I’m busy building snow plows.)

And of course, a SaaS app is rarely just a piece of software, sold as a subscription. That “Service” in the tagline matters. SaaS often involves bug fixes and new features, secure hosting, workflows, content, and other add-ons — AI agents and co-pilots are just one part of the overall offer that helps build that snow plow. SaaS is much more than an app in our PLMish universe.

So why did Mr. Nadella say that SaaS is dead? No idea. Office 365 and Azure (both SaaS) make up more than half of Microsoft’s revenue. That’s a lot of money — more than $60 billon from Office 365 alone, according to some estimates. Microsoft definitely does not want SaaS to be dead, but does want us to start thinking about how Office 365 fits into our overall workflows and enterprise architecture.

I don’t think SaaS is dead at all. The proliferation of AI means we’ll create more software to do more tasks, rather than seeing a decline. It means we’ll see new capabilities, features, and functions we haven’t even imagined yet — and not necessarily from only the biggest, platform-iest companies on the planet, since AI is a great leveler for software development. And it means our PLMish world will continue to flourish because domain expertise matters more than ever — to create the apps, train the AIs, tie together modern workflows that use today’s technologies rather than re-create old manual or paper-based workflows. Nope, not dead.