Build vs. Buy When it Comes to AI

Courtney Baker: [00:00:00] There’s an age old debate in tech that’s been around longer than clippy himself. Should you build or should you buy. Historically, that’s meant custom code versus vendor software, but in the age of ai, the question suddenly has higher stakes and a lot few easy answers. Build in-house and you risk blowing your budget on talent data pipelines in a quick pilot that drags on for a year, buy off the shelf and you could end up getting something that doesn’t really work for your business without a lot of customization anyway.

So what’s a leader like you to do? Stick around and find out. Hi, I’m Courtney Baker, and this is AI Knowhow from Knownwell, Helping you reimagine your business in the AI era. As always, I’m joined by Knownwell, CEO, David DeWolf, and Chief Product and Technology Officer Mohan Route. But first, David DeWolf joins me this week to break down some of the latest in AI news.

David DeWolf joins us [00:01:00] again. David, you can’t get rid of us, uh, to break down.

David DeWolf: get rid of me. Yeah. And I, I’ve been thinking about this because, uh, Pete is so good at the news, right? He’s like the play-by-play guy. He’s like the news reporter, like very traditional. I feel like I don’t do a good job filling his shoes because I’m more the editorial color commentator, and so I’m really trying to put my best Pete on today.

We’ll see how we do. I, I think you’re gonna get more opinion than news, but we’ll try, we’ll give it a

Courtney Baker: Yes. Well, I think this is a good one for a little bit of both,

so.

David DeWolf: do it then.

Courtney Baker: Let’s talk about the scale of investment we’re seeing in ai. Citigroup just projected that big tech will spend over 2.8 trillion. On AI infrastructure in 2029, that is a staggering number.

I feel like when we talk about these numbers, we need one of those graphs with like, you know, pieces of sand and, you know, fills a stad, some insane number.

[00:02:00] What should executives take from a forecast like that?

David DeWolf: Well, I mean, it is a staggering number. Um, and, you know, put it in perspective. That’s how we like to start here. What is global? GDP Global GDP is over a hundred trillion dollars, 110, $115 trillion. So. You know, we’re talking 3% of JDP, uh, roughly speaking. That’s a staggering number regardless, but it’s not massive.

Right? Um, and I think we expect more and more of that. that said, trillions of spend sends a very clear message that we are past the point of experimentation. This is now foundational. Um, it rivals some of the biggest. Spends in the economy that are out there. And so it is something to take note,

If we compare AI to some of the things we’ve talked about in the past around. How the internet, for example, and, and specifically internet companies exploded [00:03:00] at, and, and how they exploded when broadband infrastructure was laid out. This is that type of cycle all over again, and we’re seeing that type of scale, that type of reach, and we can all relate, like so many of us can remember the pre-internet days, right?

And how. happened to us and how the world looks so dramatically different. That’s the type of step change that we are taking here. But there’s also the nuance lesson of that as well, which is the infrastructure alone won’t deliver the results. The real differentiator is going to be how effectively do companies pair this fundamental investment in AI with human capital?

Right? We hear a lot of times about human in the loop, right? So how does the knowledge work of ai. Intersect with human ingenuity and how do we work together in collaboration? We talk about this [00:04:00] ag agentic workforce. What does that agentic workforce look like where agents and humans are working in collaboration with one another.

It’s a really critical thing for us to be thinking about. When we get to this point of scale where we’re literally spending trillions and trillions of dollars, right? Um, this won’t, this can’t, and it shouldn’t be technology for its own sake. It’s gotta integrate into our businesses. Um, and so for me, that’s the fundamental takeaway is we’re now at a point where we have to figure out the future of work.

We have to figure out how does AI get adopted in our businesses in a way that doesn’t just push forward. The economics that we’re all pursuing, but in a way that elevates and helps us take human work to the next level. Um, that allows it to be just as an even more dignifying, more fulfilling, all of those types of things so that we engage not just the agentic worker, but the actual human worker.

And we remember [00:05:00] that these businesses, we as humans aren’t here to serve them. They’re here to serve us. And I think that’s an important consideration for leaders to really be thinking about.

Courtney Baker: Yeah. That’s so helpful. And just as maybe a, a note of good news, just to remember that that projection was for 2029. And so still got time to ramp up to be proactively thinking about these things and getting your workforce ready. David, thank you as always. Okay.

David DeWolf: court.

Courtney Baker: When it comes to ai, should you build or buy? What does the research say and what do David and Mohan think? I sat down with them recently to find out

David Mohan, welcome back. Excited to talk with you two today about kind of an age old question to get your sage advice.

David DeWolf: podcast.

Why age old?

Courtney Baker: [00:06:00] Well, you know, you two, I’m not gonna get into ageism,

I’m just kidding.

Speaker 5: Wow.

Welcome

to the

David DeWolf: show, Mohan.

Mohan Rao: I feel so welcomed.

David DeWolf: Yeah,

Courtney Baker: I know.

Listen, that’s not it. You two are so wise. Um, but it’s, I do wanna bring up kind of a classic debate in technology, and you two have been in the technology game for a hot minute, so.

David DeWolf: this is just getting worse and worse. You guys have been around for a few waves of technology, so.

Courtney Baker: No,

you have such great advice for us. So here’s the deal. The question is, should you build in-house or should you buy from a vendor? Ai, obviously that question has been around for a long time, but AI makes this debate even more complicated and kind of higher stakes. Uh, according to summaries of a recent MIT [00:07:00] study, two out of three projects that specialize in AI providers are successful, while only about a third of in-house AI initiatives deliver expected results.

that’s pretty interesting research. What do you think that means for leaders making these decisions right now? Today?

David DeWolf: So I, you know, Courtney, I was, I was at a conference yesterday where the, the, this research came up and, um, there was a leader from one of the large, uh, advisory consulting firms, uh, that works with Fortune 500 companies and, and, and middle market private equity backed companies. Um, on

AI strategy.

It was the leader of their AI practice.

And

she was sharing about some

of the trends of what’s going on and talking about how one of the major problems is there was such a push for experimentation to learn [00:08:00] that people are doing a lot of these. Very tactical things that will never drive ROI. And what you see is as organizations are shifting from just experimentations for education, where there’ll never be an ROI to real.

Corporate imperatives picking real business problems, which we’ve talked a lot, but meaty ones with a real theoretical ROI that they can then go prove out that that is a big part of what is driving the successful work that’s being done. Right.

So you look at

JP Morgan is, is a great example, I think has been in the news and, and all of the work they’ve done on really

Incredible

transformation about leveraging this technology. Um, and there are other big organizations, uh, that have done it. They’re not doing little experimentations. Right. And, and the,

what this [00:09:00] lady called citizen experimentation is one thing and it’s good and it helps get the juices flowing, but it doesn’t drive the big projects that are showing up in these reports right now.

Courtney Baker: Can I just pause there for a second because Mohan. I feel like we could have saved these people some time if they only would’ve been listening to this podcast and you specifically, ’cause I know you told them to only experiment where there’s specifically a problem. And so we could have saved these people a lot of grief.

Um, so.

Mohan Rao: so.

true. Um, you know,

David DeWolf: Yeah,

just listen to Mohan. I’ve learned that in my career.

Mohan Rao: You know.

know,

Courtney, today I am gonna pull a David and say to you that this is not a good frame to look through.

Uh, it is, it, it,

Courtney Baker: Okay.

Speaker 5: He’s learning.

Courtney,

he is

Courtney Baker: the monsters I have created here.

Great. Okay.

Mohan Rao: so I think in the AI world, right? I mean, yes. I mean, there are [00:10:00] some principles around buy versus build that absolutely still stay true. But the key question is what should you do to compound your intelligence over time? Right. It is how do you create this learning organization? How do machines and humans work together to create better outcomes?

So it

should really be about, um, yeah, I mean, it, it goes

down to by versus

build, but it’s about what can you do to compound the overall intelligence, over time, sooner.

Because sometimes

building in-house can feel empowering, but it comes with.

Crazy

hidden costs. Uh, right. So around data infrastructure, governance and all of this stuff, buying gets you the value faster, assuming that you find a good vendor, but it introduces dependencies, that you have to carefully own.

So, but ultimately it’s about figuring out

how do you compound intelligence in this whole, uh, process of whether you buy or build.

David DeWolf: Well, and, and that goes back [00:11:00] to

this is where this all ties together. It’s the problem you’re trying to solve, right? If you are going and saying, okay, my AI strategy is gonna be buy versus build what? Hold on, time out. That’s like saying, let’s just go experiment for experimentation’s sake, right? The

question is, what problem are we trying to solve, right?

That leads you to what intelligence

do you

have to compound, right? And then you

can say. Okay. Now how do we do that best? And from use case to use case, there’s gonna be a different answer because as Mohan just articulated, there’s pros and there’s cons to each approach, right? With building it your own, you get bespoke.

It’s specific to you. There’s downsides of having to maintain that. All of the things with buying, get it off the shelf. You get to share somebody else’s investment, they move it forward for you, like these, all these things, but it can’t be tailored specifically for your business. Right. [00:12:00] And so I. I think we start at the wrong place a lot, and I love the fact that Mohan just reframed this for us because that, that is the answer to the question.

There isn’t a good or bad, both are good depending on the situation. Start with the business problem. It’s what the research is showing now.

And

it’s what we have learned from those legacy years. You talk about over

and over again for the last several decades,

what do we, what do we

always go back to?

You gotta start with the business problem, not the technology.

Mohan Rao: Exactly. And then you know, the business problems also come in various in varieties, right? There are some business problems that you think is core to you. You probably want to keep it a little closer, even if you, even if you buy, uh, uh. But there are some business problems that is not in the core thing, but still very important.

That clearly should be, um, a buy and not a belt, right? So. So it gets into this sort of a matrix that you’ve gotta develop and say, what’s the business problem? Is it important [00:13:00] enough? Um, is this something that I really kind of want to own in-house, invest, build that I feel compelled to build it

myself? Or is

this important, but I wanna buy this?

And

then you

figure out the hybrid combinations there, which kind of gives you the overall answer between buy versus build.

Courtney Baker: Does,

does it come into the Matrix at all this new research about the likelihood of success with a vendor versus internal?

Mohan Rao: Uh, no question about it. And this is where, um, you know, you gotta figure out is this a problem that you want to invest in solving yourself or not? Because as vendors, what we do is we are focused on one giant problem, and that is what we do every day. We wake up with that problem. We solve it every day in our company or any of the, uh, AI vendor companies.

Whereas when you’re running an enterprise, you’ve got to solve 25 different problems and you just can’t give attention to it. And that’s why I think most projects fail. I think you have to pick the right problem that you [00:14:00] feel compelled to solve it no matter what.

David DeWolf: I

mean, I, I, I’ll give a very real example of this, and I, I always hate to do this, but I, I think it’s just so compelling. We actually have to, is, it’s shocking to me every time we get on and, and folks want to see Knownwell and we give them a demo for years, if not decades, people have been trying to figure out how to measure.

They’re commercial health, the health of their commercial relationships and firms have all sort of processes, right? Most people are doing red, yellow, green on a spreadsheet. They’ve got NPS scores, csat. They know that’s deficient. Even when we speak to the most advanced that have all these data feeds and they’re measuring these 57 things and they’ve created a rubric.

We get on and we demo the platform of what we’re able to measure and the objective signals that we surface, and people are

blown away.

They’re like, oh my gosh. Like this is what we’ve been trying to do for years. Like we, [00:15:00] we, how did you do this? Right? And it speaks to what you’re talking about, Mohan. It’s because This is the mission of the company.

This is what we’re doing, and we’re not just learning from your data, right? We’ve done substantial market research. We’ve literally talked to hundreds and hundreds and hundreds of customers in your space, right? We have decades of experience doing it ourselves and failing, and we’re thinking about it differently, and we’re able to just solve that problem better than.

Anybody we’ve ever seen in the space. Right. And, and I think even though it sounds like a, a self pitch, it’s an example of what you can do when you are building a platform and a product to solve a very deep, complex problem versus if you’re a business and this is one aspect of what you’re doing.

Right.

And it’s just, it’s just very real. And I’ve seen it real time so often that it has become clear to [00:16:00] me.

Courtney Baker: Well, I think this is a really interesting conversation, I think for everybody listening. Um, don’t start with why, as our friend Simon would say, start with the problem, um, when it comes to figuring out if you’re gonna buy or

Speaker 2: Oh.

Courtney Baker: David Mohan, thank you as always.

Mohan Rao: Thanks,

Courtney. Thanks David.

Speaker 5: Thanks

guys.

Courtney Baker: By

Simon and I very tight. If you need me to text him, I’ll lemme know. You know

Mohan Rao: You, you really like, uh, Simon Sin, You cannot refer, to him all the time.

It’s

is from,

Courtney Baker: like him.

Mohan Rao: is,

is from Nashville?

Courtney Baker: No.

Speaker 5: No. Thompson Square. No.

Mohan Rao: Nashville, UK.

Courtney Baker: Okay.

Courtney Baker: The new era of Commercial Intelligence for professional service firms, it’s already here if you want a clear view of your client relationships.

And the confidence to make smarter, faster business [00:17:00] decisions. See what the Knownwell platform can do for you. go to Knownwell dot com slash 30 days to see your intelligence on the Knownwell platform and try it out for 30 days for free. Knownwell dot com slash 30 days.

Courtney Baker: If you’ve been a longtime listener to the show, you know that as this time of year rolls around, we often like to make some predictions all the way back in December of 2023. That was episode 13. David made some bold predictions about what he thought would happen in 2024.

Bad news. First, this nugget was included in a New York Times story titled The AI Prompt That Could End The World More Bad. News is 2025, not [00:18:00] 2024. Hey, David, happy Monday.

David DeWolf: What’s up, Courtney? Uh, all I have to say is I might be off a year. I’d rather be accurate and untimely than totally off base and just make, making things up. And as far as I’m aware, I’m the only one from those hot take episodes, uh, that predict the future that’s ever been called back because now I think this is the third time that the news has validated my prediction.

So you can, you can dwell on the year. I’m gonna dwell on the accuracy.

Courtney Baker: You know, I cannot confirm nor deny any of those things that you just

said.

David DeWolf: cannot or you do not want to. Which

Courtney Baker: you know, that’s not important. Let’s move on to the subject at

hand.

So just in case everybody missed it, um, here is the good news.

If the world ends, David, because of ai, you can actually die happy knowing, like you said, you were right about one thing [00:19:00] because it seems like your prediction about

a Moores law.

David DeWolf: that. I was only right about one thing, Courtney. I’m not, I’m not sure that’s accurate, but.

Courtney Baker: We gotta keep your head from getting too, we keep you humble, uh, about Moores Law for ai. It is indeed coming true from the article. The model evaluation and threat research group has conducted research that shows AI is getting better at longer and longer task, doubling their capabilities every seven months or so with newer models showing a doubling time of every four months.

David DeWolf: Now can I just point out just for a moment that the world does not have to end in order for this to be true. So don’t blame me for the end of the world and let’s, let’s, let’s uncouple these things, right? It may have been packaged in that as a proof point that just tells you how real the fact is that these models are getting better and better, [00:20:00] faster and faster.

Cheaper and cheaper. Will I say? Um, and. It’s all happening in very short timelines. I mean, seven months. Um, we are experiencing what we first experienced with chips back in the nineties and two thousands. Absolutely. Moore’s law is coming true for these LLMs and it’s, it’s really powerful.

Courtney Baker: It’s so interesting and I’m, I’m curious for everybody listening. David, if you could break down what that will mean for professional service leadership.

David DeWolf: Yeah. Well, you know, when I, when I think about the practical aspects of that, I think what it means, and I think we’ve seen this with these deep thinking models in particular, that are becoming now kind of the default standard, which is that they’re able to do more and more complex tasks, right? It’s one thing to a blog, it’s another to have multidimensional.

Inference, guide the [00:21:00] execution of the LLM itself to do the next level of analysis and then get results and do something else. And we start to think about all this hierarchical intelligence. That’s guiding the next set of instructions for itself, right? And when you start to go there, you start to see this agentic world and how it can come to be.

So apply that to professional services. In a world where knowledge work is the work of professional services, I think it means more and more that we’ll truly be able to empower people, um, with tools that really accelerate a lot of the work that they do. Um, now I would caution. I do not think it means it will totally disrupt.

You know, a lot of people are prognosticating, professional services is going away. The bespoke nature of professional services and the truly creative part of the work is not replicable by machines, right? These machines are only predicting the next token based on all of the data that is digested, and so we’re [00:22:00] getting what we’ve always gotten.

In a summarized form, not true ingenuity, not really the breakthrough idea, not really the connecting of dots that have never been connected before. what this does is it allows us to deploy these really, really smart people from professional services on the truly human aspects of their job. The real creative ingenuity of finding those breakthroughs, connecting those dots that have never been connected before.

Courtney Baker: It’s so interesting and going back to what you said about ent, I’ve personally been using. Perplexity comment. I

don’t know if you’ve tried this out. It’s so cool. It really can. It brings to life like some really amazing things that I can ask it to go do for me, and it does it. The problem is it takes so long, I literally feel like I’m in the dial up era.

I’m like

having to like sit it off to the side, but a lot of times, listen. I’m an elder [00:23:00] millennial. I’ve moved on to 400 things. I forget about the thing

I asked it to do before it can get it done and,

David DeWolf: I’ve also seen it get confused and stuck in a recursive loop that it

Courtney Baker: yeah.

David DeWolf: beyond,

Courtney Baker: Yeah.

David DeWolf: um, I was trying to do something and, and leverage it to automate something in, in Google Drive, and it just could not figure out Google

Courtney Baker: Yeah.

David DeWolf: the most basic

Courtney Baker: Yeah, but

the, the light is there at the end of the tunnel. You’re

like, it will get faster and it will eventually be

like, uh, cranking away. And so even though I am, am I sad that you got this right? I’m not, I’m not sad that you got this right. I’m actually quite glad that you’re prediction came true, but maybe I just reiterate as we close that you were a year off.

Okay. Okay.

David DeWolf: I’ll take it. I’ll take it. Courtney.

Courtney Baker: Okay.

David DeWolf: know when you’re two years off of one and we can debate if it really matters.

Courtney Baker: You know, honestly, I don’t even remember mine to know. I will ask our podcast team to pull some of [00:24:00] mine to see. Maybe I was right. You don’t know. Um, David, thank you as always.

David DeWolf: So much fun. Thank you.

Courtney Baker: Thanks as always for listening and watching. Don’t forget to give us a writing on your podcast Player of Choice. As always, we’d really appreciate it if you would share this show with someone you know would enjoy it. At the end of every episode, we like to ask one of our a i Friends to weigh in on the topic at hand.

Hey, chat, GBT. Really curious to hear your thoughts on the topic of this week’s episode. ~Should our listeners build, should our listeners build or buy when it comes to ai? ~Should our listeners build or buy when it comes to enterprise ai?

Elizabeth: That’s a great one — building can give you control and a competitive edge, but it’s also a massive time and talent investment. Buying, on the other hand, gets you moving fast but can box you into someone else’s roadmap.

Courtney Baker: Now you are in the know. Thanks as always for listening. We’ll see you next week with more AI applications, discussions, and [00:25:00] experts.