AI Agent & Copilot Podcast: Michael Simms on Why AI Projects Need Data Readiness, Strong Governance

John Siefert (0:06)

Hey, everybody, welcome back to another episode of the Cloud Wars AI Podcast, where I get a chance to sit down with different individuals in the market, talk about, hey, what are some of the opportunities, what are the impacts and what are those outcomes that we can start to expect as we employ AI within the context of our businesses. Now, today I'm joined by my friend Mike Simms. Mike is no stranger to shoot in the breeze with me on some of these types of discussions, nor a stranger to things like the AI Summit that we run in the spring every year, where Mike's been a part of the Programming Committee Board or Community Summit North America, our large gathering of individuals from the Microsoft ecosystem to discuss the innovation, education, and training needed to make it through some of these big changes we're all living through.

So Mike, thanks for taking a little time to hang out with me, my friend.

Michael Simms (1:00)

Thanks, John. I really appreciate the time here today.

John Siefert (1:02)

Absolutely. So let's, as they say, drop into this wave nice and quick because, but when you think about how much evolution we have seen in the AI category just over the last couple of years, but before even the last six months, it's been pretty daunting.

I think many organizations right now are trying to figure out the adoption of an adaptation to things like agentic workflows and all that type of stuff inside their organizations. What do you think are the top two to three things that companies are really trying to figure out right now?

Michael Simms (1:42)

Yeah, that's a really good question. There's a lot of things out there, and I think the big thing that companies are trying to figure out right now is value.

How do I get the value out of these things?

What does this particular initiative mean to me? It doesn't really matter what part of AI you're looking at, whether it's agentic or it's automation or it's machine learning or what it is. It is just a factor of how do I get value out of it. We've seen this with the numbers, and the session that I'm doing at Summit, it talks about the roadmap to get there and produce value. The statistics are pretty daunting. I can't remember off the top of my head, but it's something like in the 80 percent of initiatives are not seeing the value that are anticipated. And so that begs the question, are you choosing the wrong initiatives? Are you using the correct data? Have you defined the processes correctly? All the underlying pillars that are required for these things to actually work and actually provide value, those need to be revisited and looked at. And it's like, okay, what do we need to do to support this initiative?

John Siefert (2:46)

Yeah, I think it's interesting, Mike, you bring up a point there that we should double click on a little bit because, and I want to get to some of the data stuff because I know you've got some great experience on the data side and the readiness for AI and that kind of thing. I want to explore that in a moment. But I think your earlier point of really understanding something that to me feels almost like an instinctual thing to think about. But for many, they're really struggling with it right now. And I just actually had a conversation with some folks about this topic. But folks are going down the path of, hey, it's 1-800-AI-everything, right? You know, hey, this sort of thing over here, okay, let's build an agent for it. This sort of thing over here, okay, let's use Copilot, let's use Cloud, let's use Grok, let's use something like that to do this. But I think now what folks are starting to realize is, boy, it's not 1-800-AI, right? It's actually, hey, if I need reasoning models around decision making and workflows and that kind of stuff, boy, that's the right time to create an agent. But if I just need things like consistency in workflows and such, that's a lot better for automations, right? Just being able to automate some of those processes. But it's starting with an understanding of what the processes are and which ones demand reasoning and which ones just define consistency, right?

Michael Simms (4:11)

Yeah, absolutely. And how do I invest in these things and which tools do I use? The tooling discussion typically, at least in my book, doesn't come in until we figure those types of things out, right? Because if you start throwing a tool at it just because you're a vendor, and I want to throw this XY tool at it, let's pick on Grok, for instance.