Read the transcript below:

I think the traditional answer is that artificial intelligence is a pretty broad discipline. Computational design, in particular, involves mimicking certain patterns, workflows, or behaviors that a human might do—using, for lack of a better word, algorithms.

When you look at the history of artificial intelligence, it’s been around for a long time—even going back to Turing and the Turing Test, which asked, “What could artificial intelligence actually look like?” So the concept isn’t new.

But in the context of our industry, a good way to break it down is by thinking about the different forms or “flavors” of AI—how it shows up in workflows, how it’s applied. In the past, the underlying technologies were things like machine learning models, where you take a bunch of data and try to extrapolate patterns. There was also natural language processing, which does something similar, but focuses on understanding semantic language blocks.

Most of these systems are powered by a type of architecture called neural networks—interconnected nodes or “thinking blocks” that influence each other. I actually worked with neural networks back in high school, so it’s definitely been around a while—though I may have just dated myself.

The big leap, though, came with the rise of large language models (LLMs)—and now, increasingly, small language models. With LLMs, we’ve been able to take in much larger context windows and broader datasets—effectively, the entirety of written English history. These models generate output based on prompts that feel like they were created by a real human—something that thinks and connects ideas the way we do.

Where is it going? I don’t know for sure. But I do think one thing is certain: everything will become AI-native, or at least AI-influenced, in some form. It’s a bit like what happened with the internet. At first, companies said, “We’re the first internet-powered software.” Then came mobile and social. Now, it’s just assumed—it’s all software. Same with AI. We say “AI-native” now, but eventually, it’ll just be software.

I don’t love making broad proclamations about why something is exciting, because that’s personal. For me, healthcare is exciting for very personal reasons—because of my family and the people around me. It’s played a critical role in my life. Being able to see how software and now AI can improve outcomes—that’s what gives me passion.

If you zoom out, healthcare has long been one of the largest sources of data in the world. But historically, that data was seen as trash—maybe not literally, but at least incredibly hard to work with. That’s why LLMs are so exciting. For the first time, we can synthesize this “messy” data at speed and scale, with accuracy and context. The data generated by clinicians, scribes, medical assistants—it can now be made useful, quickly, in ways that support real decision-making.

Early in Keebler’s journey, we shared the Gartner analytics framework. It outlines how analytics can evolve from descriptive (telling you what’s happening), to diagnostic (why it’s happening), to predictive (what will happen), and finally to prescriptive (what you should do). Historically, getting to that top tier—prescriptive—required highly specialized machine learning or NLP models built for very narrow use cases.

But today, we’re seeing a shift. Thanks to language models, we can go from raw, messy healthcare data to prescriptive analytics much faster. The models are more flexible and more broadly applicable. You no longer need a custom-built model for every single workflow.

That’s the power of this moment. If you’re able to leverage AI-native infrastructure, you can reach that prescriptive level of insight at unprecedented speed. And that’s where we’re going to see some major wins.