Santiago Suarez Ordoñez on Accelerating Enterprise AI Adoption

Hello and welcome to the latest edition of the Techstrong AI Leadership Insights series. I'm your host, Mike Bazar. io, and it goes by the name Santi for short.

But we're gonna have a little chat about where are we with the state of AI and the enterprise. 'cause I think we're entering maybe the, what they call the trough of disillusionment. But we'll see.

Santi, welcome to Sha. Hi Mike. Thanks for having me.

Excited to chat about this. What's going on here? It seems like early on every executive on the planet was like, oh, we gotta invest in this tomorrow.

We gotta drive all this AI stuff and we're gonna get these incredible returns. And there's magic in the air. And now it looks like people are starting to figure out that, well, it takes work to get this thing to happen.

And there's a lot of data management issues and a lot of things that we have ignored for decades are now raising their ugly head. But what's your assessment of what's going on here? Um, personally, based on my slice of the market and the day today, we live, and I feel like we may today be on the early innings where interest is continuing to build up and, um, buyers continue to come into the scene with excitement and strong mandates to understand and spend and explore in ai.

I do agree with you that there are some indications of burn where buyers have, uh, gone ahead of themselves and spend money on things that are not really delivering. But I do think the wave is still forming there. Of course, we can, we can dive in and and debate whether we think it's gonna crash, uh, and we think there's something on the other side that's positive.

But in general, I would say today I see energy building up, uh, not quite yet deflating. I'm not entirely sure that it's a crash as much as maybe a more realistic set of expectations. And to that end, everybody I talk to has multiple, if not tens, maybe even a hundred or more experimentations going on.

But I wonder if we're experimenting too much and maybe we should just pick three or four things that we're actually gonna get done and put into a production environment. I think you're right. I think one of the interesting qualities of this new, you know, disrupting function that is ai, this new type of technology that is driving such transformation throughout all industries and all solutions, is this, um, density to make every single demo work every single first take on a new solution.

An AI forward attempt to whatever problem that may be, you know, affecting a buyer tends to work pretty well, at least for a demo on first try. So I think that's created an explosion of cool demos and superficial solutions that buyers can go spend their money and time on. Um, I think one of the biggest challenges that buyers have today is to really be thoughtful and, uh, double click to understand what is there beyond that first sexy demo.

How operationalized a workflow is for the long run, how scalable a solution can be for a big team like an enterprise company. Uh, I think that's really what's at the, at the center of not getting burned, spending a ton of exploratory money on AI and then having to turn up a bunch of vendors, uh, to a month down the line. Mm-hmm.

Are there any patterns you're seeing in use cases that are maybe the first low hanging fruit that people can really operationalize? And maybe everybody should because everybody soon will, but at least I'll have something to show for my efforts. Uh, yes.

I think ulti, at least of course this is, uh, talking in my book, right. I do think, um, data extraction, using AI, driving a data set that is inherently unstructured and requires a bunch of manual admin work to have the business be able to leverage at scale and having LLMs be the conduit through which that becomes structured, um, is really good. It's a really good use case.

It may not be the sexiest one, right? You know, what is sexy today in the industry? Well, what's sexy to it in the industry is to come and pitch you that I can replace 2000 people with, you know, an LM for you.

And that's kind of the easiest thing to go try and sell. 'cause everybody would be in the market to dramatically lower cost. Um, but as it turns out, that is a, a really hard thing to deliver on in the long run.

There's a lot more nuance to solving those types of problems. Uh, something that may not sound as sexy, uh, but it's more concrete and more contained of a value prop to say, Hey, I'm gonna grab something that is very time consuming and gives you low quality data for the business and turn it into more reliable, higher quality. I'm not going to clean a bunch of people's jobs, but I'm going to dramatically reduce the amount of admin they have to do so they can focus on something better.

Uh, that's a little bit of what we sell ourselves every day. Yeah. It seems like to your point, we got a little obsessed about the labor arbitrage equation when maybe we should just be thinking about, well, people are spending a third of their day on tasks that are either manual or intensive and toil that doesn't really add a whole lot of value into the business.

And if we get rid of that, well then we'll get more value out of the people we have. Correct. That is a thing.

What's going on? Um, usually what you see is, um, founders and startups built on the idea of only replacing labor and those, and to be younger founders, we not really lived in the enterprise who don't understand the complexities of a big organization and the new ones of having humans as the glue between complex processes and communication at chains. Uh, so they come and they idea idealistically say, well, if AI can write an email, then I can replace a team of 1500 scr with a bunch of ai.

Uh, but then when you put those in practice, uh, they crash and burn when they hit the hard reality of, you know, the real world. Um, what you're seeing is more, um, senior, um, uh, veterans from the field will come and put very concrete solutions. They can slot it into a big enterprise for a very simple problem, and those deliver quite well.

Uh, so that's what I would kind of nudge towards every time I would talk about, okay, how does these get operationalized? How does it, how is it integrated into my tools? How is it blended into my processes?

Uh, that's the real questions to be asking in, in 2025 when there's so many vendors, there's so many submissions around Mm-hmm. To your point, it's not clear to me that organizations have a big appetite for large scale business process re-engineering, and they kind of want some of this stuff to slide into their existing workflows and processes, because otherwise it would just be too disruptive. But it does seem maybe on an evolutionary scale, we will re-engineer these processes.

It's just gonna take a while. Exactly. Exactly.

It's the same, you know, I have an analogy too on companies that today are rebuilding the CRM from the ground up and say, Hey, look, we are a, we're a brand new AI forward CRM now who is gonna go dump their, you know, Salesforce or Microsoft Dynamics since this overnight, when you have a team of 2000 sales reps who have been using it for, you know, 11 years to capture all their data? Well, not too many companies, they'll find a lot of small startups that may, you know, take a spin on a new approach. But the majority of the industry will not mitigate impact.

And the companies that are, you know, leading the transformation are the ones that are making the current situation incrementally better. I may end up leading to a destination that is the same, but you know, may take a couple of years to get there rather than doing it overnight. Mm-hmm.

One of the things I do hear people struggling with is they'll take a notion and they'll be like, wow, we can use AI to create this, and they'll even maybe get as far as creating an AI agent and then they'll wake up the next morning and one of their vendors already did the same thing, and we will give it to them as part of the application that they're already using. So where do I find that line between when am I gonna build versus buy something that, um, I can add some unique value around versus reinventing a wheel that somebody else is gonna do for me? I mean, I'm pro buy versus build every time.

I just feel like if you are in a business that is thriving, that is growing, you have a market that is hungry for your solution, it is rarely the case that you should be spending your time building some workflow optimization for operationalizing your company. Uh, there's somebody who's, you know, there's some vendor out there whose only focus is to do that, right? And you will never be able to do it at the degree of refinement they do it.

Um, it usually tends to cost a lot less than it takes to build, uh, even in the age of AI with agents around, you know, upkeep, maintenance, uh, continue to be considerations. So in the build versus buy, uh, and a spectrum, I'm a big fan of, you know, buy while you're growing. Sure.

If your company is, you know, growing 10% or year over year and you've kind of tapped to the bulk of your total adjustable market, and now it's about optimizing cost, sure. Then go consolidate, then go build your own business from the ground up. Uh, but if you're thriving and your business is doubling year over year or tripling year over year, I wanna get my ass off the ball.

I would just focus on my market, uh, with every piece of technical resource I have at my disposal. Mm-hmm. We're humans though, and we all get obsessed with quote unquote making things better.

And I think if I buy, I wanna customize that 'cause um, we're all convinced that we have unique business processes and that may be debatable, but, uh, people are people and that's what they want do. So do I also need to figure out how to choose platforms that give me something that I can buy, but I can extend Completely. I think in the world of ai, what you're referring to is the idea of prompting what makes AI custom, what makes AI behave the way you really need is to be able to highly refine the prompt that is ultimately given to the model.

Even having the selection of what model is being used gives you a, a degree of customization and control on the end result that I think is really deal breaker between something that's kind of cool for a demo, but not really practical in the real world and something that really works the way a business wants it. So in my book, it's all about integration automation and prompting that will be at the core of a solution that feels your own versus something that is vanilla and doesn't really get you to level of, uh, replacement of labor that you expected in the first place. Mm-hmm.

One of the challenges I hear also is that, um, most of the business processes today that we're trying to do are very deterministic in that sense that they're supposed to be done the same way every time. And the one thing AI never does is the same thing the same way twice. Um, so we have a probabilistic set of technologies that we're trying to insert sometimes into deterministic workflows.

So how do we do that and, and how do we strike that balance because well, you know, it's, it's neither a hundred percent of one way or the other. Yeah, this is a really good point. I mean, I think the oversimplification of some of these patterns and tools into simple categories and words, things like agents, uh, can lead to issues like the one you described ultimately, you know, if you really were to go to the very core of what an agent really should be, you know, the word agency comes up, right?

An agent needs to have agency of what it's doing, and therefore it becomes a very istic, non-deterministic, um, type of execution you would get out of this technology. Um, and I don't think most enterprises out there are looking for something like that. In reality, what you're, they're looking for is basically a codifiable deterministic workflow that has a certain level of intelligence when it really matters.

You want something to be 90% programmed and always behave the exact same way and then be 10% intelligent in the parts where, you know, simple if conditionals won't cut it. Uh, that to me is really what most people talk about when they say they're buying agents or they're selling agents, uh, real, real agentic behavior that is fully null programmed and probabilistic. I'm rarely seeing it in the field, and I rarely see buyers interested in bringing that in.

Mm-hmm. So what's your best advice to folks about how to get started with all this? I think we have some on the one end irrational exuberance.

And on the other end we've got people who are probably overall and too terrified to do much anything at this point. So between those two extremes, how do I get to something that feels like a reasonable middle? So I will say everybody should be bold and everybody should be hang hungry for this.

You should not be waiting this one out. If you are, your competitor is gonna be exploring it, and if they do, they're gonna be ahead just the delta in results from the people who get leverage from Gen AI today versus the ones who don't. It's just too big.

Uh, so I do wanna drive urgency on, on companies out there no matter what you're selling or what you're building to go use it every day. Um, you know, we put together a book and, you know, regardless of whether buyers, uh, whether you're listeners get it or not, the book is called Ignite, uh, go Market with ai. I got to interview a lot of interesting people to put together the book, and I'm gonna steal this one from Kyle, or he's the c of owner.

com is a thriving company, uh, hundreds of reps, uh, performing really, really well utilizing ai. And he gave me one of the most interesting insights and approaches to this question. He said, look, Santi, the companies that are gonna thrive the most are the ones who have their leadership be AI forward.

I think an anti-pattern that is developing today is to have AI adoption come from the bottom up. Let the front lines explore ai, understand AI and bubble it up to leadership from leadership to decide what gets bought and what gets used. Uh, I do agree with Kyle that the right pattern here is to have your CEO, your CRO, your CTO, be the ones hungry for innovation, be the ones understanding the differences between Quad and Chad, GPT, understand what an embedding says, what a vector database is, and if they do, then they're gonna be the ones making the most bold decisions they wanna be, they're gonna be the ones buying the most disruptive technology.

Um, so, so that's my advice. I'm copying Kyle and saying, you better have a CEO who understands what this tech can do. You better have A-C-R-O-A-C-T-O-A-C-P-O that are looking into this technology and are spending an hour or two a week dabbing in into the latest and greatest 'cause that will drive the best results at the enterprise.

Ultimately, is there something that you are seeing organizations do that just makes you shake your head a little bit and go, folks, we could be a little bit smarter than that? Um, I think organizations are quite exploratory today and they're very boldly bringing in vendors and trying to stack, and I don't think that's a bad idea. Um, I think where it really can be, uh, a bad approach is when you're being careless about data these vendors handle for you and how they use it.

I think buyers should be looking into their MSAs into terms of service to understand if data is being used to refine models. If data is being used to train models, uh, you wanna be in business. If you're a business, you wanna be in business with vendors whose only interest is to charge you money to provide you a service, and the only reason they're gonna use your data is to provide that service.

And whenever the engagement is over, your data is yours and it gets deleted on a, I agreed upon I'm, um, time commitment. So that would be my one component is don't be too careless about your own business data these days. Data is a commodity.

It's, it's something that the companies are looking for, it's a currency, uh, so you should protect it and be responsible about it. All right, folks. You heard in here one way to think about ai.

It's an undiscovered country, and the only wrong decision was to stand still and do nothing. Hey, Santic, thanks for being on the show. You bad, Mike, thank you for having me.

And thank you all for watching the latest episode of the Textron Do AI Leadership Insight series. You can find this episode and others on our website. We invite you to check them all out.

Until then, we'll see you next time.