Practical AI

Don't Read Too Much into the Collapse of Applications

• By Howard M. Cohen
• 02/25/2025

The original title for this column was going to be "Don't Read Too Much into How AI Will Change Us," but it occurred to me that might be pandering to people's fears too much, and people would definitely read too much into it. What I want to talk about definitely will change us (people), but only in some very specific ways having to do with how we encounter, interact with, and use software.

Reducing Applications to CRUD
One of my personal favorite terms, CRUD. Stands for "create, read, update, delete" and it is considered to be the four things any application must be able to do. This notion supports the concept that any application has two fundamental components. The first is simply a database—no particular database, just some flavor of a database. The other is the business logic, according to which the application will "CRUD" the data.

Consider this from your own perspective. If asked how to describe how you use software, you'd probably include steps like running the software, getting to the main menu, selecting which activity you want to perform from the main menu, providing some specifics or criteria, and then letting the application process what you've requested. You anticipate an output that informs you or tells you what to do next in the physical world. Or some similar permutation of that.

Don't You Love It When a Plan Comes Together?
In recent articles here in "Practical AI" I have reported on two things that, when taken together, form the foundation of what Microsoft CEO Satya Nadella spoke about in a recent podcast. One of those articles was AI Changes Our Relationship with Applications and the other was Welcome to the Agentic World.

Nadella was far more blunt about his point, speaking of business applications in general and saying, "That's where they'll all collapse. In the agent era."

Yes, the CEO of Microsoft is predicting the collapse of business application software. He then said, "The business logic is all going to these agents."

Let's be sure we're all clear here about how truly monumental this prediction is, especially given who it's coming from. Essentially, Nadella is saying that since the beginning of computing, whenever you think that is, we have been using application software in pretty much the same way. We've been given a user interface (UI) through which we can tell the software what we want it to do. The software then interacts directly with whatever databases it uses. It seems like no big deal, right? It's the way we've always done it.

It looks like it's time to start thinking about changing that, because Nadella is predicting that the choices we've been making are destined to go away. Instead, we'll simply be telling an agent what we want it to do with the data and the agent will do it. The agents may interact with each other to get certain things done, but the interface to us inferior humans doesn't go any further than typed or spoken chat. For anything and everything we need done, there'll be an agent for that, and we'll ask the agent to do it.

All of which raises the question: How will you be preparing your people for that major change? That question is well worth thinking about and planning around.

Agent Efficiency
There's a suggestion here of what I would call human augmentation by agent. Nadella explained, "The biggest benefit is the agent efficiency, right? Where the agent is happier, the customer is happier, and our costs are going down."

Although you could probably lose a lot of sleep worrying about what he means by "Where the agent is happier," his core message is that we humans will have agents assist us in getting everything done because they are much more efficient than traditional applications. They're not replacing us; they're replacing traditional software. If you want to assuage that worry and get a better night's sleep, repeat that phrase to yourself over and over. (I do.)

In my recent conversations with "typical users," the very first question they asked, consistently, was "Where am I going to learn how to program these agents?" My response usually underscored efficiencies, and I pointed to Microsoft's Copilot Studio, "a tool that lets you create AI-powered agents to automate tasks and interact with users." A typical user of this tool tells Studio what it wants an agent to do, and they describe it in plain, natural language. Studio may ask some qualifying questions, but ultimately, the user has done no coding, no programming in the traditional sense. They've simply asked for something that does what they want or need done.

Fundamental Changes
Clearly this has implications for almost everybody. Typical computer users will have to change the fundamental way they work with computers. Businesses may be freed of a great deal of cost but will need to have strategies to assure their users are asking for the right things. Anyone who has anything to do with software application development had better consider insinuating themselves into the technologies behind AI. Even those regular users who have embraced "low-code/no-code" platforms may find themselves abandoning them.

And the implications go far beyond these few examples, which is why I'm asking every one of you who reads this column to respond in comments with your thoughts, your observations, your considerations, and your concerns. Perhaps even solutions if you have them. It's a dialogue we need to have, because this is clearly a "yellow woods," a place where two paths diverge.