Why Most AI Initiatives Fail with Paul Shirer

Hi, I'm Michael Hartman. I'm Naomi Liu.

And I'm Mike Rizzo. And this is OpsCast, a podcast for marketing ops pros and rev ops pros created by the MoPros, the number one community for marketing operations professionals. Tune into each episode as we chat with real professionals to help elevate you and your marketing operations career.

Hello everyone. Welcome to another episode of OpsCast brought to you by marketingops.com, powered by all the MoPros out there. I am your host, Michael Hartmann, joined by no one again. We'll get Mike and Naomi back at some point here, I'm sure. But today I am talking with Paul Scherer, founder and CEO of Infinite Ideas AI and director of AI and go-to-market technology at Bridge Partners. Paul spends his time helping organizations evaluate, implement, and scale AI initiatives. And in earlier conversations with him, we found ourselves coming back to a theme that many ops professionals are wrestling with right now. That is, despite massive investment in AI, many organizations are struggling to generate meaningful business impact, that is measurement. We'll explore why that is, where companies are getting AI adoption wrong, how leaders should be thinking about workflow design and decision-making, and what practical steps organizations can take to move beyond experimentation and toward measurable value. So Paul, welcome to the show. Thanks for joining.

Hi, Michael. Yeah, thanks so much for having me. Yeah, excited to talk today.

Yeah, it's an all-Texas affair today since you and I are, what, about 50 miles apart? Is that about right?

Yeah, that's about right. You got the north side covered, I got the south side. We got our territories we stay in, so.

There you go. There we go. Yeah. Well, again, appreciate it. So let's just dive right in. So when When you've talked and what we've sort of traded messages about, you've said that many organizations are treating AI as a tooling problem, when in reality it's a workflow and adoption problem. So, break that down. What does that mean in your words, and how did you get to that sort of point of view?

Yeah. Well, I got to that point of view by failing, number one. So, that helped over a little time.

Pain is a good teacher, right?

That's right. Um, yeah, so I remember, I recall, you know, back, whatever, a couple years ago, and maybe it was 2 and a half years ago now when the ChatGPT moment hit. And, uh, the first thing was like, we need to get ChatGPT, right? And so we got it. Everyone got a license, you know, and it was like, and then you'd go look at the usage of that tool and it would be like, you know, very few people using it at the time. Now I'm sure that would be different if we did it today, but at the time it was very little.

Yeah.

And so you start looking at, uh, you know, uh, how are we going to start transforming the way we work? And really that was the theme from the beginning. If you were getting ahead of it from a strategic standpoint, you were thinking this is going to change, but how long is it going to take? What's it going to take to do that? And what are the baby steps along the way? Right. And so a tool is just a tool. It's sort of like a hammer is just a hammer. Like we hear that all the time.

That's right.

What it is. But unless you have something to use it on and kind of what you're trying to build, then it doesn't, it's not really all that useful. So I think that the key to this whole thing is, for most organizations, is, you know, start where people are. Don't get way, way ahead of yourself. You know, like the idea that you're going to change away all of your workflows and processes and get people to follow along and do that all at the same time. It's a very difficult thing to do. And I think a lot of people, a lot of folks have done that in terms of how they've piloted AI. They're trying to rewire everything. And so I kind of advocate for meet people where they are, how they're currently working, get them to buy in, get them to actually start to use these things and see the value and let them be your research. You know, and one of the metaphors I use, and not such a metaphor, but sort of strategic framing I use is we're working inside organizations, but it's not much different than if you were going to market with a new tool. And so think about your organization as the marketplace, if you will.

Yep.

And what are the pain points? What are the problems, their challenges that they have? And how can we have this new technology come in and help solve some of those problems? So meeting them where they're at, you know, getting their sort of buy-in in the early early stages of it? How can you help them create better content, make better decisions, you know, look at analytics in a different way, all the things that we know AI can do, but talk to them about it and start to begin to understand how you can support them with the tool itself. And that means, you know, obviously skilling, it obviously means, you know, onboarding and creating communities of practice, make it be their idea, so to speak.

Right.

Help people share those ideas. Anyway, point being that I think a lot of folks have been trying to solve the whole thing all at once. I don't think that that's reasonable in terms of transformation within an organization, and you have to have a strategic approach to looking at the workflows, looking at the adoption first, and then kind of slowly moving towards how can you kind of impact and operationalize AI with a specific workflow.

Yeah, I mean, I like that idea. So for the audience who's listening only, I'm sitting here nodding my head all the way along through Paul's description there, because I think that I like that idea of thinking about this as a little bit of an internal sales and marketing initiative, right? Or go-to-market initiative. That's an interesting one. But your point about meeting people where they are, I think is an often missed one with kind of all change management. I can't remember if we talked about this when we last spoke, but I did some research on change management as I was working with a client a couple years ago and I came across a Harvard Business Review article from like, I want to say it was the late '60s. It may have been the '70s. I don't remember exactly when it had been revised, but what struck me about it was even though it was that old, right? The principles were still there. And I think, you know, this idea that one size fits all is not realistic. I think, and I don't know if this is true or not, I would imagine there's sort of a bell curve, right? Of people who are like, on the one end, you've got people who are going to be like jumping right in. They're hot to do this. They want to change this. On the other end are the people who are like totally reticent. Why break something that's already working? And then like over time, like you've got this whole big chunk of people who are sort of somewhere in between.

Yeah.

And so thinking that they're all on the one end, right, is probably not going to be successful if you treat everything that way. Now you could take advantage of those people. I think that you were hinting at that, right? You were talking about, pockets of community. I mean, if I think about like that pocket of community is the ones that are going to really test the boundaries and then they can help advocate and teach others.

Yeah, exactly. I mean, I think a good skilling program, there's different levels. We actually put together kind of different levels of skilling program with one of our, as we were initially rolling out this about a year and a half ago or so, that we Really spent a lot of cycles on this. And the skilling is like giving them some free courses that are on YouTube or LinkedIn Learning. That was obviously the obvious or low-hanging fruit. And then some certificate programs, maybe paying for a few things. But one of the best things you can do is identify those champions and those folks that are the super users that, and really they can train the folks. They're happy to share all the practice. In fact, they're very vocal about it.

It's an ego boost thing for them, right?

So yeah, and it's a career path for them in some ways too. They're seeing it as an opportunity to build not only their skills in their position, but also maybe even leap within the company or who knows, go somewhere else. But sure, they will be the leaders of a community of practice. So we've stood up probably about, I guess, 3 or 4 communities of practice now within companies. And, you know, you just get them started in Slack or Teams and Have those folks that really want to be the champions come in and they're very vibrant internal communities. I was actually surprised at how vibrant they would be.

Yeah. I mean, I've seen the same thing internally. I mean, I'm a big fan of like the internal communities for things like that. I think the only thing I've seen that has hampered them is when there's too much concern about what might be said inside those, right? It's kind of the same with it is when you have a, say, customer community. But like, I'm always, I tend to go like, let it go. Like, and if it happens, right? First off, you're, I think it's like the, like teaching kids, right? They fear doing something new and it's usually the fear is way greater than the reality, right?

That's right.

And I think I'm all for like, let it flow.

Right.

Don't, I mean, set guidelines about the way that you can communicate that's acceptable. So that's a little bit different, but if like, in terms of the content itself, if they say this tool sucks, right, let them say it, ask why, right? And what I found is that then the rest of the community who doesn't agree will counter it.

Yeah, it's kind of like social media.

Yeah. I mean, well, yeah, I mean, hopefully with some more guardrails than what you see. So, all right, well, so another thing you and I talked about is that you, and I see this too, right? I've seen companies, I've seen comments about people sort of accumulating tools, agents, spending sort of unexpected large amounts of money on tokens, things like that with some of these tools. So when you use the word agent sprawl, right? How do you think about that versus maybe a more appropriate adoption of AI, that it might be more useful for organization?

Yeah, it's a great question. And, you know, not having a perfect answer by any means, but just observing what I've seen in kind of the evolution. You know, we first saw, we just talked about the idea of having tools was the first thing, and then prompt techniques were everywhere, right? And everybody was sharing those prompt techniques. And That was kind of a nice, fun little age where people were just playing with these things. And then of course agents come around and what, just to kind of define what agents are for this point here I'm making is the—

Sure.

The Copilot agents within Microsoft 365 or the ChatGPTs within OpenAI or Gemini Gems and things like that. Now there are some other, a little more robust agents that folks were building, but by and large, the citizen-developed agents that are in play that are out there and I don't want to, I'll take it a little further. There were folks paying for, you know, more production level and for development companies to build some agents too. But the problem with it all was it was all done kind of in a vacuum and all done in terms of, we need this now, let's build it and let's, we're not going to think about how everything connects together. You know, what my co— my colleague is doing or what this other department over here is doing. Not going to think about, um, you know, all the governance or any of these things. We're just going to do it. In part, that was also because leaders were saying, use AI IRLs. They were freeing up budgets to invest in AI and they just said, go.

Yeah.

And so what we ended up having was all these agents. Um, some of them were duplicate efforts. A lot of them were lower quality citizen-developed agents. Um, things that were, you know, helping like product marketing or sales outreach or content development or analysis. And they've been doing some things, And, but they by and large weren't being used in the most efficient ways, in the most productive ways, and there was just a lot of waste. So that's what I call that agent sprawl factor.

Yeah.

We saw this a lot at enterprise, but I think it was happening pretty much everywhere, even in midsize, et cetera. But that's the agent sprawl, but I'm happy to go further and talk about, you know, What, why that happened and what we can do about it and what's the next evolution, if you like, as well.

Yeah. I mean, I think I definitely see it. I don't think it's isolated to, in fact, there's a part of me that thinks that it's probably less of an issue in enterprise organizations because they're so hard to move. There's a part of me that thinks it's like in like early stage small companies where people sort of, they just figure shit out and how to get things done, right? They're going to be adopted with less of that 4-letter word governance, right? Yeah. So I'm curious, I mean, are you seeing some of that more in enterprise organizations or are you seeing more of it in mid to smaller companies?

It's definitely both. I agree with your point on the, you know, midsize is the kind of the one where it's the most propensity for enough people to do some things, but less governance. So that's where it's happening, probably the biggest problem area. Smaller, there's just not enough people to mess it up anyway, so you're kind of—

Yeah, yeah.

And the implications are—

Yeah, the risk is lower in terms of like the scope. Yeah, I get it.

Yeah, yeah. And then, but enterprise, yeah, definitely saw it. Now they have, and I work with Microsoft, Google, SAP, AWS, so we're talking about big cloud enterprise.

Sure.

There's a couple other hyperscalers we work with, That kind of thing and their partners. But now what I saw there, what I've— we're in the midst of is there was a whole fix my agent phase. So they citizen developed and then we came in and fixed their agent, so to speak.

Yeah.

Now, um, so they were seeing, and then our responsible AI started really coming down the pipe, right? And so in particular on the Microsoft area, uh, responsible AI is a huge part now of The overall, just getting an agent into their system now is becoming a big thing. But, um, the fix my agent was the thing, uh, for a little while. The biggest thing that's happening now is, um, that's this stakeholder and this one business unit over here is talking to this stakeholder because they have teams and matrixes, uh, you know, across all these different areas.

Yeah, sure.

They're very similar in a lot of ways. Microsoft's maybe a little more matrix-oriented. But they're now going, wait, you're doing this, I'm doing this, you're doing this. Maybe we can bring these things together. So they're getting pretty proactive about bringing things together and not just working in their silos.

Yeah, that makes sense. I mean, that's the one that, it's the benefit of having a larger organization. You probably, if you think about that bell curve I mentioned, right? There's going to be just more volume of people at any one particular place in terms of their propensity to use the tools. So, curious, you know, and I think you and I, when we talked about like, you're thinking about evolving, you're thinking about AI and how to incorporate it, implementing it, and all that has evolved over time. So if I remember right, you felt pretty strongly that you need to wire together end-to-end kind of automated workflows, call-up agents, whatever, But it sounds like you're, I think in our conversation, it sounds like your perspective has shifted. So maybe talk about like, yeah, what was your belief kind of early and then how has it evolved and what, like, what is it that's changed in that process? Yeah.

Yeah. Thanks for that question. And then that is really probably at the center of my thinking that I'm really focusing on now with the teams that I work with as well as The reality of the concept of an end-to-end workflow where you can, first of all, just agree on the key steps of that workflow. Say it's a sales prospecting all the way down to nurture, down to some level of conversion.

Sure.

Um, or a content pipeline or whatever it may be, especially in mid-market. Um, the idea that there are quote unquote key steps and micro tasks that happen that are rinse and repeatable all along that cycle is a little more illusion than reality. You know, like it's, you want it to be true. Theoretically, it sounds like it can be true, but then you get in to do the work and you find out everyone's different. You find out there's nuance to it and there's like customization all the time that's necessary.

Shadow processes, shadow systems, right? Yeah.

Yeah, exactly. You got all these different things that you weren't aware of and you You always, and this goes back to all the days that we, we tried to be, to engineer workflows, right? We've been trying to be agile for years, or not just that, not scratch that word agile, or I don't really mean, I meant it in the kind of the—

Not the general term, not, not the methodology. I get it. Okay.

Yeah. The, when we first started talking about workflows, this was our land. But the point is, uh, that we, um, we've been trying to do this for a long time. And so we thought AI would be the forcing function in a way to get our stuff together so that we could have the right AI doing things and human in the loop or human on the loop And we can construct these things. But the struggle was not only just that, can we get something that is rinse and repeatable, but then getting human beings to be in the loop.

Yeah.

It's like rallying cats, right? To come in and do this thing and to believe and to buy into it and all that stuff. Now, I'm not saying I gave up on that. I do think there's certain pieces of the puzzle or the pipeline where you can come in and get some pretty nice wired workflows. And, you know, having AI autonomously do some pieces of the puzzle that would be called agentic workflows. I think that stuff is real.

Yeah.

But yeah, I just, I just decided that, you know, the bigger opportunity and the thing that we kind of just rushed right past is just getting people to be better at using the tools that they have and expanding out the workspaces, which I won't go into that right now. I think we're going Probably talk about that a little bit more, but that is what I've seen. I know it's contrary to everything you're seeing on LinkedIn. I get it. They're all workflow first. And I was writing about workflow first a year ago. I was talking to my leaders in companies, but as we've tried to do it, I just, it just hasn't always played out the way we expected it to.

I just want to make sure I'm clear. So when I hear you talk about that, I'm thinking like a workflow, like say it's, Let's say you can even agree on, like your point about agreeing on what the steps are, right? But let's say you've got general agreement, you've got 10 steps, and you try to automate the entire process. Is that what you're talking about? Like, okay.

Yeah, that's a good question. Yeah, I think it's, when I look at workflows, it's always gonna be some mix of what AI can do, what automation can be done, and then humans take a piece of the puzzle and then they're in the loop. But it requires this official formal design of a workflow. And that's, you know, this conflict. So I've seen a lot of, and it's, again, I was talking about this over a year ago and I'm taking back just a little bit, is that a lot of folks have said you have to invert. And that's what I was saying, you know, back then too, is like invert from people first to workflow first. And so instead of Hiring people, and this has implications for how you change your workforce and who you hire and all that kind of stuff.

Absolutely.

Hiring for the workflow versus having people do it and then have them use tools. So it's just, and, and again, I haven't given up on it, Michael. I, I do think that there are some spots where this is useful, but I think that this sort of blanket statement around workflow-first mentality is, is, is not as practical as it, as it sounds.

Yeah. And the reason I asked you that, I remember seeing LinkedIn posts when I was still sort of, I felt like I was late to dipping my toes into using tools like ChatGPT or Claude or whatever on an individual basis and kind of thinking about it as I was working on a problem or some sort of deliverable, right? It wasn't my normal, it took a while for that to become a part of how I thought about getting things done. But what I remember seeing is some people who had tried to do, whether it was, you call it agentic workflows or whatever, that had like lots of sort of AI components in an overall automated workflow, right? Maybe there was some human-in-the-loop component to it. And it, like, people were sharing like, this is what I'm doing and here's what it looks like. And then it felt like it wasn't that long after when I started to feel like, actually what we find is If something goes off the rails in one step, right, it has this cascading effect. Yeah, to me, it's like, I think about like, okay, if I'm— this is going to maybe offend some people. Like, if you're shooting a gun at a target, right, and the farther the target is away, right, any little variance is going to mean at the end when you get to that target, it's going to be way off. And it feels like that's kind of what was happening is like people were like, oh, I'm going to automate this whole thing. But then something would break here. And actually what I think I remember is like, not only did it break somewhere, it was harder to figure out where something went off the rails.

Yeah. Yeah. It's, I don't, again, we would probably need 15 different use cases to really break this. And, you know, if you're doing data consumption and you're extracting data and turning it into some organized form, and then that's gen— manifesting an email that gets pushed to your business leaders, maybe that's a useful automation that probably would work most of the time. If you're having AI make decisions for you based on some level of criteria, I think your example comes into play more. It's like, yeah, that could go off the rails. That's not— and then, you know, the AI for all its promise, and I love it. Look, I use AI every day. I'm a— this is what I do for a living.

Sure.

You know, the generative AI portions are always, they're just not quite right yet all the time. And so we can't just have it do things for us, you know, and do it in these automated ways. But as you add human-in-the-loop components to it, the biggest problem we have there is then, so let's say that you have an end-to-end, I'll just do a content engine real quickly. We'll just do a real campaign-in-a-box concept real quick.

Got it.

So a campaign in a box, you would have a creative brief of what the campaign is. Um, you're probably marketing a specific product or service. You have target personas. You have probably a message you can position, your framework that governs that, brand guidelines, all these things that can help you kind of put together a content campaign. And then maybe your content campaign has an ebook and then maybe has a one-pager and then has nurture, it has reach-out emails, nurture streams, ads, all these things that are kind of in that campaign that help that Sure. Well, if you put together an end-to-end workflow that says, we're going to pipeline this thing, we're going to say AI's going to do this part, humans are going to do this part, AI's going to come next and blah, blah, blah. You put this whole thing together and you say, got it. We got this really streamlined thing. Our efficiencies are going to go down. We still have quality gates. I mean, all these things are great. And then you, then you go put that into practice. Well, you've got writer X over here, That has to pick it up after this? Where did it go? Where's the project management tool we're gonna use? What's the status of this thing? Well, now do we have to put a project management tool in this thing? Okay. And, and people have done that. And then it's like, what's next after that? What's next after that? And so what you thought was this theoretically easy end-to-end thing, it started to get, it started to get really complicated when there was new writers coming in. Oh, that person was fired. No, they're not here anymore.

Oh, that.

It's like, you know, there's just the idea that it's just going to somehow work, you know, that's the problem.

But yeah, you try. Yeah. And so you hinted at this a few minutes ago about the idea or concept of, I think you call it connected workspaces. So rather in that as an alternative. So first off, maybe like, that you did earlier, like what's the definition of a connected workspace? How's it different than some of these other scenarios we talked about where things seem to be, appear to be connected, but aren't really, right? They're disconnected. And then what does that look like? Do you have an example or something where you can talk about it in the practice? Maybe what you just went through is a variance.

Yeah, I think we've been evolving to this in general with different solutions of late, and I'll try to lay this out without having like visuals and, kind of get looking at these examples, I think it might be a little hard to follow, but I'll do the best I can to kind of lay this out. Um, we've talked about some of the, um, underlying problems, the, the agent sprawl, for example, right?

Mm-hmm.

Think of a marketing team or a sales team, um, and, or any go-to-market team, right? And they are, uh, working, they, they started off, they got these tools, they're starting to work in silos, they're building all these, and we got the agent sprawl issue. Um, you then go, okay, well, why don't we create workflows that will, again, we just talked about the workflows. These are all part and parcel where let's end to end. We're going to have agent X come in here, human in the loop do Y and things like that. So that was another sort of address, way to address the kind of the work to be done in that area. And we've already talked about some of those problems. So now what we have is like co-workspaces that are growing up. So popularity of Claude and Cowork, and Microsoft has its own word, um, its own Cowork space, and we'll see more and more of these. But these are really personal workspaces still where people are doing work and they're doing it in all their one-off ways in their own idiosyncratic personal ways, which is what they should do. But what I think is emerging is you still have marketing teams, you still have sales teams, and you even have teams within those teams that do work together. And so, I'm, what I'm seeing is this concept of a connected workspace where it's built around a specific team's needs. And it's anchored in all the data that they work together with, the documentation they work with, the processes too. But—

Shared context.

It's a shared context. And so, one example we've been building out recently has been around product marketing teams. That also, of course, do activation campaigns. They also do sales enablement and things like that. So you get it, you get at a crossroads and you go, well, aren't there some tools out there that we can just buy? And then you go down the buy world and you go like, well, no, not quite right. It doesn't quite fit me. It's not exactly what it was. And then, then what do we do with all of our data? Like, is that just all over there? Like, what are we doing with it? So what I'm seeing is, and we're building out a couple of these recently, they've been really resonating with clients. And so we're happy. That's happening. That's always good.

Market validation is always good. Yeah.

Yeah. So, and so this is kind of an exciting area where we're building all, all those agents that work on a disconnected, or they were just plugged into Microsoft 365, or they were GPTs or whatever those. Instead, we're building one space where all the product marketing needs are there, where you can build the vision and mission statement for the product, the, um, the overviews, the features and benefits, the ICPs, All there, but it gets stored in a data layer. So it's not just knowledge sources that are unstructured, like all these documents that are you putting in there.

Okay.

We're storing it in a data layer. And then they go to create blogs or they create ebooks and things like that. And that's all getting stored in the data layer. And so are all the chat sessions. So what we're creating is really this completely connected space around a business function where all the rich data's being built up over time. And so when you go to each agent to build something, whether you're building the blog or you're building a draft of a, of an email sequence and all that kind of stuff, instead of having to reground every one of those agents in the documents, it's pulling from the same data layer across everything. And, uh, it's creating a much more powerful way to do work and do it at scale. And I think that, I don't know where this all will go. I'm just saying what I see, Michael, and what we're building.

I swear, like I've given up trying to predict, 'cause it's like, it feels like every time I turn around, there's some new thing out there. And if I go back to, like we talked about, like the adoption at different types of companies, I mean, I know of at least one organization that I've heard where developed their own sort of tool for people to use, you know, LLM kinds of tool based on some of the other models. And then using IP restrictions, like blocked all the other ones. Well, new ones are popping up all the time for specific use cases and people are getting around it or they turn off a VPN or whatever, right? I mean, that's the stuff that's really happening out there.

Yeah, the shadow AI, I've actually got a post that's going to go out in the next few days. I've been doing a research piece around it first, and I'm going to post around it on shadow AI and what to do about that. And I think that we're seeing a, first it's like, don't ask, don't tell. And then it's like, let's be punitive. Let's like, oh, well, you know, do all that. But then I think the next phase, which I believe we need to look at very closely, is let's let that be a tell on the market. What do people want? Because these are the people that want to use things. And so ask them, you know, embrace it, intake what they're doing, and then maybe you could put together a plan about which tools you're going to buy to support them. Even though shadow AI is still going to be there, at least you're embracing what's happening and know that that's not nefarious. It's like, I just want to do my work better. I want to move faster.

Yeah.

And to embrace it, in other words.

Yeah, no, I mean, I'm again kind of going back to this, like I would rather have that be in the open and know about it than just think it's not happening, 'cause, you know, and then people feel like they can't talk about it. I think that's a worse kind of outcome. So you talked about a data layer, and I want to maybe dive into that a little bit, 'cause I'm not sure this is exactly what you were talking about. So if not, that's fine. We can kind of connect the dots here. But yeah, one of the challenges I think everyone in ops would tell you is like, our data sucks in go-to-market, right? Yeah. Whether it's marketing, sales, whatever, especially in complex B2B sales stuff. How, 2-part question, because I have a strong, so how important do you think that data quality is to enabling these kinds of tools? And then one thing that I would like to see, I haven't heard a lot of people talking about is I'd like to see AI being used to help improve the data quality and completeness. So in other words, don't wait for the data to be whatever, pick the level you want, right? Before you start using AI tools for other things, but like, how can the AI tools help expose and correct the data? Like, what's your take on all that?

Yeah, no, that's a great question, Michael. And I think to just say it and state the data's the problem is one thing, but to then unravel like what's inside all of that as you work through it, And I think, you know, it's a problem in so many ways and it's always been a problem. It's just that AI is now exposing it because as you try to really leverage the power of AI, which essentially it's like it just needs data, it just loves data. It can't do anything without data really to really be effective or to be differentiating or useful. So you just could go across any area of the go-to-market space, you know, all the way up and down the customer journey, if you will. And you're gonna find, well, you can't personalize outreach if you don't have data, right?

Yeah.

You can't follow, you can't have on-point battle cards and sales enablement if you don't have data. I mean, you can wing it as much as you want, but you're gonna need that intake. And so the struggle we have, and You know, is that data, uh, in an unstructured form is not all that useful. It has to be structured and we know that. But, um, so how do you go about doing that is the question. Um, and how do you pull from all these sources? Well, um, sometimes, um, it's just a matter of like, you can't even get access to a lot of data. Like, you know, sometimes it's not collecting the right data. Think of all the—

Sure.

Just the entire trail. We're not collecting all the data properly. There's that. But sometimes even within organizations, it's just so many systems, whether it's within Salesforce or, you know, different CRMs, um, or maybe it's all your ad data over here. Yeah. Your customer intelligence or feedback loops, the market research that you do, that's over here. And, um, so we're like, just recently I've been working on, um, competitive intelligence a lot for whatever reason. I have sold, uh, 2 competitive intelligence solutions. And it's really rearing its ugly head around like, well, we have external data sources we can grab with AI, which that's available on a lot of levels.

Sure.

But where are all your internal sources? Like, well, there's documents just gathered everywhere. There's CRM data, there's sales forecasting data, there's our customer feedback data, and it's everywhere. But the challenge is then, How do you get access to it, pull it together, normalize it so it can actually be used, and then you can actually have AI do all kinds of levels of RAG responses and, you know, ingestion of it and do things with it.

Right.

But it's just, it's a long process and it's the thing that, and the biggest thing I would say is this, and like if you go macro, sometimes we can dissect these things, but if you want to just put it as a big equation, Data's super important. Leadership doesn't want to invest in it because it's not, they don't understand it. It's not sexy. It doesn't make sense.

Right, right. The immediate, there's not an immediate obvious value to putting in the effort and cost. Yeah, totally.

They don't want to invest in it, but it's super important. So crappy output. It's the simplest equation in the world. But, you know, I think that the good thing is that AI has exposed a lot of things, and we talk about this a lot, exposed our workflow processes issues. We've exposed, you know, talent and skilling issues, has exposed data issues. And I think that's one of the best sort of value benefits we've gotten out of AI.

Yeah, I think, I actually too, like, I still, I think I still feel like there's a play here to use AI to help with that data quality. I just don't, I don't know What it would look like, you know what I mean? Yeah.

Well, if I agree, no, no, AI, that's one of its beautiful things it can do is it can organize, right? It's taxonomy.

Yeah.

Though I, it's, it's so much better than us and it's so much faster. And so I agree with you on that. In fact, we were doing a project, there's another project we're working on where it's a first thing we're doing is we're just doing some data auditing. And we want to build some dashboards for OKRs, you know, the objectives and key results for a quarter. And so the first thing we're doing is we're going through like 15 different dashboards that have already been created, which is like nuts. And—

Dashboard is a 4-letter word to me.

Yeah, it is at this point, you know, and if you do it right to some degree, but the problem is people never go— there's a lot of Yeah. Anyway, nonetheless, the, we can use AI to go into these data systems, which, you know, some, the inconsistencies of data across all those systems, those dashboards. Sure.

Of course.

We are using AI to go in and extract all that and create a master taxonomy, like a sensible master taxonomy that can come all of that in order to then build the, the next dashboard. Um, but then we'll also have push, you know, notification, you know, push information and things like that will come off of it. But that's one way AI can just intelligently sift through it and suggest a better taxonomy. That's so simple.

Yeah. Well, I mean, I'm sure somebody's got to be doing this, right? A very common thing for an ops world, both marketing and sales ops probably, is get an external list, you got to normalize country or state or whatever, right? And I'm like, that's a, or Normalize job levels, right? Based on titles.

Yeah.

I mean, I think that's a powerful way of using it, right? Doing that organization for you and doing it dynamically, right? So over time, like you adjust based on what the real data is.

Yeah.

As flawed as it may be.

Yeah, absolutely. I think that's a great point. And the, well, I'll just say that my, the basis of the business I'm building, Infinite Ideas AI, is all based on the idea that AI can create new and unique data in the market.

Right.

And so that's, yeah, my, not only agreeing with you, like, it's like, to me, I was looking at the way we, all the signals that are out there in the world, you know, not just around conversations on Twitter or Reddit and things like that, and sort of the typical sentiment analysis and things like that, but just taking the everyday sort of things that are happening out there and turning them into real signals. You know, what's the velocity of investment in a specific space? What's the, what's the, if you knew what your current product was and you, and you saw something happening, some news item happening in the market, how related is this news item to my product line and should I do something about it? And you could score that, right? Right.

Like a risk assessment or an opportunity assessment. Yeah.

Yeah.

Yeah. I like that. It's interesting, your point. Like, this is the other thing that really, it wasn't until I was talking to someone who was a writer at one point and I, we were talking about something else and I asked, threw in the question like, oh, what do you like? 'Cause I'm assuming most writers detest, right, these LLMs for what they do. And this person was like, well, we do breaking news so they can help us make our writing better, but they can't break new news 'cause there's nothing to go on.

That's right. That's right. I do love that part. And when people were talking about content creation first, that was kind of an area that I was like, well, you can always do new reporting. You can always do new things. You can, there's things happening everywhere and that are unique that, yeah, that LLMs are not trained on. Not yet.

Yeah. Well, I mean, there's this whole thing I saw that like, if you're a top performer in whatever you think you're gonna use AI for, Uh, the likelihood that you're gonna stay a top performer is actually pretty low. Like it's gonna bring you down to the median, right? And if you're a poor performer, poor, it's gonna help you up. Like, so—

Oh, interesting.

I mean, I was like, oh yeah, that's like, oh yeah. Like, so if you are like, if you're a top performer, should you be using AI? Well, maybe you should be using AI differently. This like, it begs a lot of interesting sort of questions that go beyond that.

Great point on that one. It's an equalizer. We used to say that when we were playing, I played a lot of basketball in my life. I still play a lot of basketball. And, when we played basketball in the rain, it's like, well, there went your speed advantage. That's gone.

Yeah.

It's like there's no first step in the rain. So, everybody was equally the same speed in there.

Yeah, yeah, yeah, yeah, for sure. Outdoor basketball or if there's a lot of wind, right? Yeah.

Sorry, man. Yeah. Football is same, right? If it's raining or snowing, it doesn't really matter who was better anymore.

Yeah. Yeah. All kinds of good sports analogies there, I'm sure. So I want to kind of take 2 ends of sort of the whole process of AI adoption, and I would like to get your thoughts on sort of both, because it feels like there's a linkage. The first is, for the people who are listening or watching this, right? How should they be thinking about whether— so if they're leaders, how should they be thinking about How to get AI initiatives started, right? What should they be doing? Maybe what should they not be doing? What are some guiding principles there? Should they be looking at doing things like building versus buying, especially if it's like, call it agentic stuff? I hesitate to even use that term. But then the one that keeps coming up regularly now is then how do they set things up so they know that they can measure The benefit from a, I think really from a financial standpoint, maybe there's non-financial benefits too, but like, what's your take on that? It's a big-ended, open-ended question for you there. Yeah.

You know, that's good, good, good questions. And I think that that's what we're, that we're in that era now. Like we had the fun, there was sort of blank checks written. It's like, we'll figure this out later. The competition's doing it. And so we gotta do it. If we don't do it, we may be left behind. Fear factors, all that stuff. And now it's like, okay, wait a minute. Where's the real ROI and how should we really be deliberate and have some sort of method to our madness and how we actually make decisions around this? Which I love, is which kind of the area I focus a lot on. And so now having an expertise in that area actually makes a difference. I don't know, it's kind of interesting because I always like look at like how many people are actually Googling or what's the trends around AI strategy or something like that, you know?

Mm-hmm.

And we see that uptick. I kind of follow that a lot being in the strategy space, and it's like, I think they're starting to get the point that we might have to think about this and that kind of thing. But so, you know, when you're looking at, if we start with, I guess, a buy versus build, I'll kind of give you a couple of anecdotes on this to kind of show you the difference and might help guide the decision.

Sure.

It's really gonna come down to, and this has always been the case in buy versus build. This is not a new thing, right?

I think I know where this is going. Okay, go.

Yeah. You know, it's like how bespoke, how bespoke to your business does it need to be? And because there is going to be a place, there's gonna be a time when it's just needs to be more bespoke. It's not like it's one size fits all.

Well, Should all.

It's just, we know that's not true. Um, even more precision, like level and more nuanced and niche-oriented solutions, it still can come up short. Um, I'll give you an example of a client I just met with yesterday, actually. Um, and it was interesting. I was glad to hear it. Um, you know, it was like, look, I can go to compete intelligence tools right now. I know I can do that. Told you I was kind of working on a couple of compete, uh, solutions lately. But It doesn't help me because it's not bespoke enough to me. It doesn't, it doesn't know, it doesn't work with the data that I really want it to work with. It doesn't deliver it the way I need to deliver it. It doesn't have the lenses or the views or the outputs in the way I need to see it. So there's just one thing after another. And so I think if you're going to do a buy versus build situation, that's the number one thing you gotta be looking at is And sometimes it's, there's nothing wrong with, this is, I remember getting this question a long time ago. It's like, there's really nothing wrong with buying now and then building later.

Right.

You know, you can gap it, you know, you can get, and you might learn from the buy experience to then maybe build later too. So sometimes a decision's just better than nothing at all.

Yeah. Well, I think just, so what I expected you to be talking about is the trade-off Right? For getting something more bespoke is that now you're taking on support and development maintenance.

You know, that's a great point. I do want to, when you're ready, I want to comment on that.

Don't. That's it. That's it.

Okay, good. Yeah. Yeah. No, that's, I love that point. And this is the crazy world we live in with vibe coding, you know, and with agentic-level coding. And it's like where you would have to buy in the past, you may be able to spin something up that's reasonable. And I mean like someone who knows what they're doing. I don't mean like go out to, uh, you know, a Lovable or Bolt and just like put up a prototype because a lot of live coding is just prototypes that don't work. But, you know, doing a lot of development, uh, with working with developers a lot, you know, with Cursor and now with Cloud Code and things like that, you can get things pretty well bespoke for sure, but you can get it on point. and in market with some level of MVP pretty quickly now. So you're, the build versus buy equation has changed because of that. I'm not saying that that's for everyone. I'm just noting that that is a, it's changed the equation.

Fair enough. True. Yeah. Yeah. No, I think that's, that's a good point, but there's still like, that is something you have to take on, right? It's not a zero cost.

Technical debt. It's technical debt that you don't know where it's going. That's the big unknown. You will never know where something you're building is going to go. And then you need good product managers. And I think that's always a way misunderstood skill set in the world.

Yeah.

Product managers know how to evaluate something for its business impact. They know how to t-shirt size things to know how big they are and how long they're going to take and how much they're going to cost. They're going to know the technical debt, but most organizations don't have that muscle, so to speak. No, I agree. They think they have it.

Yeah, I think so too. So, okay, so there's a buy versus build. Like, how are you seeing companies that are successful in measuring the impact of AI initiatives?

Yeah, another great one, right? And so you can build all the ROI calculators you want. There's a lot of them out there.

All models are inherently flawed, right?

We can vibe code one right now.

Right.

No, no, but seriously on this, you know, there are some things you can do before and after. I mean, there are clear things where like conversion rates on a sales funnel or something like that is a clear thing that can make a difference. If you've got outreach, like you've got volume and quality equations with say outreach and sales, you can get ahold of, did the quality versus the quantity game work? Was the personalization of this AI helping me in this matter? You can do obviously time to be able to make decisions. Like decision time's huge when it comes to data and how you can ingest it and things like that. Those are things that I think you can get ahold of, but a lot of it's just going to be pure qualitative. And I think a lot of people are scared of the qualitative, but I love it. I mean, I like the idea of go ask your people. I mean, what's wrong with that? You know?

Yeah. The analogy I think you and I talked about was it's sort of like attribution reporting, which is the—

Oh yeah, yeah.

And I think it's right. It's like, it's rife with all kinds of data issues and challenges with measuring and you can't capture everything. But I'm with you, right? I'm a big believer that a lot of marketers, and part of why I think they have their percept— the perception of them sometimes in organizations is not great, they forget to use the skill they should be really good at, which is storytelling within their organization.

Storytelling. There you go. That's awesome. I love that point a lot. I'm going to bank that one for sure.

Good.

Because, I mean, that's the truth. And, you know, have all these, what I'm dealing with a lot is CEOs or leaders in orgs that have to get a board-ready thing. You know, they have to, it's like, it's going to be storytelling no matter what you do here, because there is no way to say anything right now that would make everyone happy. I mean, there are investments being made. There are lots of proof of concepts. There are a lot of things that we're trying to do and understand, but you have to make the investments to to get to the next step. Otherwise, you're going to be left behind. And so it's not the thing that a board member wants to hear, or like someone who's paying and making the investment, because it's not how they want to work. And anybody just wants results, right? But the reality is this is, I've actually said it the other day, I was like, wow, yeah, you're right. Like, this is kind of a cool concept where the whole world is in an R&D phase right now.

Yeah.

We're making investments in things that We just need to figure them out and we have to kind of, we're all collectively doing it together.

Well, and I know of at least one person on LinkedIn who's been posting regularly in the last 2 months, probably. Here we are mid-June 2026 about like how these organizations are, you know, the CFOs are also seeing large chunks of money going out the door, you know, so say they're using like the Clay is the easy target here, right?

Yeah.

People are like, oh, look at all this great stuff. We can replace something like ZoomInfo and do this waterfall enrichment process. And then the bill comes in for what it costs to do that. And it's this hidden cost that I think are going to drive the— there's going to be the forcing function that's going to push people to go like, we've got to figure out some way of measuring the impact because otherwise there's just too much Too much money leaking out of the organization.

Yeah, totally. And not only that, modify behavior because of it, you know, because token costs really are becoming huge. More than the token war, or like the model war, they, I felt like they all figured it out. We'll just keep charging people and keep promoting fear that they don't get the latest model.

Yeah.

Or double. Yeah.

Anyway, well, it's been great. This would be good to like, maybe let's wrap up here and it's too bad Mike Rizzo couldn't join us because I think he would like us to ask, he would want me to ask this question if he, or he would if he was here, which is, I think you and he, I think the two of you talked before, kind of, he said that you have a similar view as him and some others that like future CMOs are going to come out of an ops space. Or at least have a background in ops. So I'm curious, like, is that true for you? And then if so, the sense I get is you think that like AI, like a lot of this AI activity is going to be something that might accelerate that process.

Yeah, yeah, yeah. Completely agree with Mike on that. And I think it's a macro thing across leadership, but it's particularly in this area. You know, leaders of the past, and I actually did some research on this because I was wondering about what's the shift in terms of technology-driven or oriented leaders versus, you know, like the classical, like maybe financially oriented or some sales-oriented leaders. Like if you look at CEOs and things like that. And it is shifting considerably and you see in a lot of key companies, You know, like the NVIDIAs and the, uh, even Microsoft when it shifted to more of a, uh, tech-oriented, they, they got their stuff together. Microsoft was becoming a joke, right? And then 7, 8 years ago they shifted and now they're, they're getting track. And I see, we see this everywhere. And I think it's the same within marketing ops and marketing. Um, I mean, how can you possibly lead an organization, right? But not having some felt or understanding, like you gotta feel things. It's like the, the, I don't know if you've seen White Man Can't Jump. It's a movie.

It's been a long time, but yes, I know what you're talking about.

Yeah.

Yeah.

It's like you, you can feel, you can't, you hear Jimi, but you can't feel Jimi when he was talking about Jimi Hendrix.

Right. Yeah. Yeah. I don't remember the theme, but that's actually a true statement, I think. Yeah.

You know, that's the, that's the difference. Like you can't go into like just hearing it and somehow you gotta feel it. You, you have to have been there. And it's always hard for me to kind of talk to people, like, and explain to them why you can intuitively understand how something's built. Well, if you built something 100 times, you just know, you just can see it and you know where it's going, kind of. You're not always right, but you got a real good sense of it.

Yeah.

And I think if you don't have that, you're really in trouble. And it's just, I can't see how it plays out. Now, there's one other piece of that puzzle. I do still think you need leadership skills. You know, like leadership skills are a thing.

Absolutely.

You gotta know how orgs work. You gotta know how to sell something internally. You gotta know how to work with people and lay the groundwork. All those things are, you know, everyone we know about leadership skills, but a powerful combination of having that marketing ops background and knowing how this thing functionally works and feeling it, plus the leadership skills, I think is the combination.

Yeah, love it. Okay, well, I'm glad we got to get into that question a little bit. So hopefully people will stick around and listen to that. That's an important piece. Hey, so Paul, again, thanks for joining us. If folks want to continue, like they're interested in continuing and going deeper on some of these topics with you, what's the best way for them to do that?

Yeah, go out to LinkedIn, Paul Shire, P-A-U-L-S-H-I-R-E-R. I'm the only Paul Shire I think out there. There was one other, he's got a financial planning company and we talked and so he helped me think very well. It's a good guy. But no, Paul Shire, I'm the, you know, literally Paul Shire. And then Infinite Ideas AI is the business I've launched and would love to even hear even feedback on that if you all get a chance to take a look at it.

Fantastic. Again, Paul, thank you. Thanks to our audience out there for continuing to support us. We are always grateful for your input and support. And if you have ideas for topics or guests or want to be a guest, you can reach out to Naomi, Mike, or me and we'd be to get the ball rolling. Till next time, bye everybody.