Why Marketing AI Initiatives Fail Before They Even Start

Episode Transcript

Here is a fun riddle for you. If two marketing teams have access to the exact same AI tools, same technology, same budget, same mandate from leadership even, why is one team three times more likely to have AI fully integrated into their work? Okay, it's not really a fun riddle at all because that gap is kind of a bummer, but the gap definitely exists. We've measured it and discovered that agile marketers are three times more likely than their non-agile counterparts to have AI fully integrated into their marketing processes.

And in the last year alone, that number jumped from 27% to 39% for agile teams, while non-agile team adoption stayed flat at 13%. The same tools, are available to both groups, so tools are not the variable. And here's another data point that made me stop. When we asked both groups what's getting in their way with AI, they named completely different problems.

Agile marketers said, "We need clearer policies to guide our ongoing usage." Non-agile marketers said, "We're worried about accuracy and quality."

In my opinion, that is not two teams at different points on the same journey. That's two teams on fundamentally different journeys. Today, I am digging into why that gap exists and, more importantly, what it' actually takes to close it, because the answer is not a better tool rollout, it's an organizational challenge.

And good news, I have got the perfect person to help me make that argument. Welcome to the Agile Marketing Edge, the first podcast dedicated to turning agile theory into real-world marketing breakthroughs. I'm Andrea Fryrear, CEO of Agile Sherpas and your guide on- this climb to smarter, faster outcome-driven marketing.

Every week, we dig into the what, who, and how behind agile marketing, from building high velocity workflows and cutting waste, to measuring what actually matters and scaling success across teams.

This week's episode is about AI adoption, specifically why it's stalling in most marketing organizations and thriving in a smaller number of them. Our data points clearly to a cultural and structural explanation rather than a technology one, and my guest today, Melissa Reeves, has spent her career building the frameworks to explain why.
Her new book, Hyperadaptive, is essentially a blueprint for what it takes to go from, "We are running some AI projects," to, "We are an AI native organization."

And to do that, she draws on lean, agile, and systems thinking. So I am thrilled to have Melissa Reeves on this week's show. Melissa, welcome to the Agile Marketing Edge. I'm so excited for our conversation today. But before we get into the data and all the super nerdy agile stuff that I know we're gonna talk about, I wanna just start off with kind of the premise of Hyperadaptive, because you've got a pretty striking claim in there that 80% of AI initiatives fail, and it's not because the tech failed.

That's right. So maybe give us a, uh, a quick overview of what's really going wrong when these AI initiatives are crashing and burning. Yeah. And of course there's, there's more than one, factor at play. Uh, you know, it's just... it's not just one thing that's at play. But I feel like one of the things that, that happens, and it's counterintuitive, is that these, these organizations are racing ahead.

So I feel like a lot of organizations see the promise of AI and agentic AI, uh, marketing organizations, other enterprises, parts of the organization, and they race ahead to orchestration, and they want that holy grail. But the reality is, is if you don't put the foundational elements in place, you often get 80% of the way there, uh, a partial solution, and now you've burned through your financial capital, you've burned through your political capital, you have people who don't really trust AI, and that's part of that 80% failure rate.

And I feel like the other big factor in that 80% is a mismatch between what organizations are trying to achieve and the AI solutions that they develop. So it might be mismatched to your business goals, or it might be a cool one-off solution that then gets abandoned, uh, in the background when people actually go, go to use it. Or it might have been vibe coded and just isn't production ready. It doesn't stand up to the test of volume or security or all the other things that a true enterprise grade solution really needs to succeed.

Yeah. That, hmm, everybody just wants the, the end, right? Like skip over the hard stuff and get to the cool stuff at the end.

Mm-hmm. Yeah. It's so true. Yeah. And you know, we do our annual State of Agile Marketing report, and we've started to probe on- the AI plus agile equation for the last couple years, and what' you're describing really matches up with what we've been seeing and I think is an interesting maybe through line of, of what we're gonna talk about. So let me share a little bit of that, and then I'd love to get your take on it. Uh, so one of the most diagnostic pieces of data that I think came out this year was when we asked marketers in particular what was getting in their way when they were trying to more successfully implement AI, which is obviously a broad phrase, but that's how we phrased it. Um, we got very different answers from agile teams and non-agile teams. So the non-agile teams are still concerned with accuracy, quality. They don't have trust that AI, is gonna do what they want it to do, or it's gonna be safe, or it's gonna introduce a lot more risk. So they've still got the skepticism barrier there. But then the agile teams were struggling with a lack of clear policies and guidelines. Hmm. So they were at the point where it was weirdly, like an absence of guardrails was- Hmm ... somehow their constraint. And so for me, it seems like these are people on two totally different roads. Like we haven't even diverged from the same road, we've just jumped to a totally different highway. Um, so when you're looking at organizations that are in that kind of messy middle that you talk about in the book, does that, this kinda gap match to what you're seeing? And, and how does it, uh, translate to the overall organizational readiness?

Yeah. I, I mean, it's interesting data that the, the State of Agile Marketing surfaced, and I d- I don't know that it's cleanly divided into like these lack of clear policies and, and guidelines versus concerns about accuracy and, and quality." I mean, what I see is actually more, uh, curiosity, psychological safety.

So, uh, a lot of what I see is, is we have a group of power users that are really leaning into AI, and, I can absolutely see how those who, um, have already embr- embraced agile ways of working have created an environment for those power users to flourish, right? So we've already embraced, uh, these ways of working that feel new. We've embraced a growth mindset.

We're not afraid of new things. We're not afraid of experimentation, and we're not, uh... Yeah, so that means we're not afraid to experiment with AI. And so then maybe I can see where the follow-on is those policies and guardrails.

We wanna make sure that we're operating within those policies and, and guardrails. And then what we're seeing is that there's a layer underneath those, and that those individuals are using AI in just a really lightweight manner.

You know, I'm, I'm drafting emails, and maybe I'm doing some lightweight analysis. And, um, and I actually think that that speaks more to capacity management.

And so I'm so full up with everything going on, that I don't even have capacity to learn this thing called AI. And so I feel like, um, that's again where the people who are practicing agile ways of working have a much better handle on their capacity. And so yeah, it's, it's interesting. I, I haven't necessarily seen that directly, but it's not surprising to me that people who are embracing agile, uh, are able to embrace AI in a much more comprehensive way.

Mm-hmm. And any kind of work management process that allows you to prioritize, and when something- Mm-hmm ... is important, like AI is important to us, we want to make sure that a successful adoption happens, then we can create the space and the capacity- Mm-hmm ... to focus on that.

Uh, otherwise, like you said, it's just running around. There's no shortage of things to do in any- Right ... given day. And so fitting, learning about AI, using AI, and then sharing with other people much less. What's working? What's not? "Hey, did you try this cool new thing?" Mm-hmm. That's just not happening. That's right. Yeah. That's really- And even what you were describing there, you know, you're, "Hey, have you tried this cool new thing?" When you are practicing agile, you've got these, these feedback loops built in.

And so yes, you might have reserved capacity to do that knowledge exchange. It might show up in a retrospective. You also have much better alignment, and so you know why you're, you're producing what you're producing, whether it's with AI or not. And so is this kind of getting into that phrase that you use about sensing and responding at AI speed, I think it is, that, that is really gonna separate the groups?

Because if they, if everybody wants to try AI, everybody sees the potential benefits, but they just can't move fast enough, they feel like they're always falling behind.

Mm-hmm. Uh, but then if you're an agile team, you've got sprint cycles, you've got the natural process of iteration. You mentioned psychological safety. There's a, uh... The question is no longer, like, how do we find time or how do we make sure this works, but how do we govern it? Mm-hmm.

Yeah. Yeah. I mean, there's a lot there. So, you know, d- in the book, I outline five durable capabilities, and it's everything from AI-augmented decision making to AI-enhanced sensing and responding to a value orientation. And when you look at those durable capabilities, a- again, they, they're very well aligned.

You and I share this, this background in agility, so it's no surprising that I've pulled forward some of those concepts.

And I, I think what you're, you're pointing at is, is just the rhythm of agile and how it supports the integrated learning loops, which is another one of these durable capabilities that I believe starts to get stronger and stronger as we, um, as we further integrate AI. Because I don't know about you, Andrea, but in some of the, the organizations I've worked with, it's one thing to have a retrospective and come up with these learnings, it's another to get them into the backlog and actually action them.

Mm-hmm And I f- I feel, and I'm hopeful that AI starts to take those learnings and integrate them back into the processes so that we can finally, uh, start to evolve and capture more of the learnings and action more of the learnings. Mm-hmm.

Yeah, getting them into the backlog, such a challenge. Doesn't feel like- Such a challenge. They just are orphaned out, there, aren't they? Yeah.

And so I think a, a lot of the listeners here are marketing leaders who are thinking about how to improve the way they work, the way their teams work, the way that marketing is perceived in the organization.

For those who might be still kind of paralyzed on this- Mm-hmm ... how do I start, right? We're talking about a process that is already in place that supports not only agile ways of working, but also AI adoption effectively.

Is there a place that you've found is a good maybe on-ramp, something they can think about to get started without having to blow up their entire operating system? Yeah. I mean, I think the thing most leaders, whether it' be a marketing leader or somebody else in the organization, like, one of the disconnects is that AI is really easy to learn. Or sorry, it's easy to use, but it's not necessarily easy to learn. And the analogy I give is a piano. Like, anybody can go up to a piano and start banging on the keys, but you can't be a concert pianist overnight. And I don't know about you, but for me, I have just spent hours and hours and nights and weekends trying to figure this thing called AI out, and it's not easy to learn, and there's a lot of nuance there. And so the, for the leader who's thinking that this will install itself, you're sorely mistaken. And I think, you know, the tie back to agile, and it's so interesting and almost ironic, is that with agile, we, we don't expect it to install itself.

We say you need support roles, you know, whether that be a formalized Scrum Master role or a PO role or a, a, a center of excellence. Whatever that is, we recognize that our human beings need support in installing these new ways of working, and I believe that AI is a new way of working.

So if you are an, a leader who is looking to get started with AI, first of all, don't expect it to s- to install itself.

Two, start thinking about the support that your human beings need. And three, start small, and start aligning that to a business outcome. So maybe it's how do we, um, how do we integrate the feedback from our campaigns more, uh, in a more solid way, just like we were talking about, these integrated learning loops. How do we create more variance on a, on a campaign?

And I really advocate for taking a look at an end-to-end workflow as a team and really thinking about how can you start to inject AI into that workflow in a meaningful way, and I feel like that's what, uh, what that starts to avoid, is what I call random acts of AI, where you just have people just using it willy-nilly, and then you don't get the return on the investment because people are just using it to no end.

And so I'm wondering about the, like, crossing the chasm ig- uh, idea, that concept here. It is the goal then to, to help everybody come along with your early adopters and your champions, to help everyone get over this, this chasm and into the place where AI can deliver ROI- Mm-hmm ... push toward real business value.

But you've gotta have a way to keep everybody kind of moving, that support that you're talking about, keep everybody moving so that the early adopters haven't run out of steam or gotten frustrated- Mm-hmm ... before everybody else caught up with them.

Yeah. Uh, you know, I think what you're alluding to is this bringing the rest of the organization along. And you know, I just finished writing a book, or not a book. I did finish writing a book.

An article about why the carrot and the stick aren't working. And so many leaders that I talk to say, "Well, you know, I, I tried to give them a couple hours to use it every week, and we, we had these, you know, these great Slack channels, and we had all these incentives for people to use AI, and they still didn't do it.

And then we moved over into the stick, uh, territory, and we said, 'We're gonna put it on your, your review,' or, you know, 'Use it or else,' and that didn't work." And I, I go back to Deming, who said it's, it's really a systems problem, and management owns the system. So have you put the structures in place to support not only the learning and usage of AI, but doing that in an ongoing way?

Because I feel like, to your point about the Geoffrey Moore and cause- uh, Crossing the Chasm, well, that works except that the goal line, you know, the, the goal line keeps moving. Right. And so that bell curve keeps moving because the models aren't standing still. So I think what we need to do is really appoint people in the organization whose role it is to keep on top of these model changes and e- empower them programmatically to spread that knowledge throughout the organization. And I feel like in that way, we, we, one, we design a system that's self-refreshing, and we design some infrastructure.

We, we task people with that job of staying on top of the models- Because that's the other part is there's incredible cognitive load of having everybody in your organization try and stay, uh, current on AI. So let's, let's task it to one or more people, have them spread that knowledge throughout the organization, and in that way bring the organization along rather than relying on the carrot and the stick. Mm-hmm.

Which is not dissimilar to like an agile transformation office- Mm-hmm ... that we might have seen when that was the, the organizational goal.

Mm-hmm. You ideally dedicated people to pushing that forward and staying on top of the transformation and things. Do you see that analogy holding up, or is there a distinction between how... A lot of our listeners will understand, right, a transformation office and have experienced that. Is there a distinction that you would make sure we understand? No, I think you're, you're on the right track. And, and when you peel back the layers of an agile transformation office, you see John Kotter and Leading Change. And so that's exactly it, is y- when I was writing Hyperadaptive, it stands on the shoulders of giants, right? It stands on the shoulders of John Kotter and Peter Senge and Learning Organizations, and Clayton Christensen and Disruptive In-Innovation.

And now I've extended it for the age of AI, saying AI is really another transformation, but unlike some of these other transformations of the past, it's, it's an ongoing effort.

And so whereas you at some point might have said that agile transformation office, like we can, uh, we can reduce the staffing because we- we're over the hump, um, I don't know where that hump is for AI.

I think it's gonna be ongoing for probably the next five or 10 years at least. It's good job security for those folks. Yeah. And so then, you know, you have some really great case studies in the book from organizations that have started to accomplish this. It's probably not ever, like you said, fully done or fully accomplished. It's an ongoing effort and an ongoing initiative for them. But can you tell us a little bit about how the most successful organizations are doing this differently and what that-Mm-hmm ... looks like? Yeah. And I, I feel like we've, you know, we've been dancing around it with the infrastructure, with the agile transformation office or AI transformation office. I t- I call it AI activation hubs. And, um, uh, Moderna is such a great example.

So the first thing they did right is they set what they called their AI North Star. And so that was to produce 15 new drugs in five years with the help of AI. And I know that you're familiar with the pharmaceutical industry.

Getting one drug out in 10 years would be a monumental accomplishment. So to do 15 new drugs in five years is, is huge.

And yet, when you think about that AI North Star, it's a rallying cry for everybody in the organization. If you're in marketing, you're like, "Okay, I c- I know what we gotta do." If you're in finance, you're like, "How do we finance that?"

If you're in legal, you're like, "Whoa, we might have to change some things in, in order to do that." And so I feel like, um, that's just one way that organizations can set themselves up for success.

And then the second thing that Moderna did is they were able to get 100% adoption of, uh, I think they were using OpenAI, so ChatGPT, in six months. So they held prompting parties to identify their top 100 power users.

They then, uh, took those power users to help spread the knowledge horizontally. And just looking at those lightweight, uh, horizontal communi- uh, coordination layers is another thing that they did very well. So if a good idea pops up in finance, how does that spread throughout the organization?

And this is the Yoko 10, uh, concept from lean manufacturing, which is the horizontal spread of information. And so you can see how all of these different concepts are out there, and I do feel like AI is the forcing function that smooshes them all together and says, "Here's, here's all these things that all these people have been talking about for years," and the leading organization already had the seeds of them and have been able to capitalize them to really take advantage of this moment.

Mm-hmm. And, you know, we haven't talked a ton about the cultural or sort of mindset things that need to be in place there, but I can imagine that it has to be broadly okay- Mm-hmm ... to feel like, okay, I'm gonna ask a stupid question here, and nobody's gonna judge me, or I'm gonna push back on a process or a workflow- Mm-hmm ... that's always been kind of the sacred cow of the organization, because this is the thing that I believe is gonna impede my function from contributing to this North Star. So allowing people the, the psychological safety to say those hard things, to feel like it's okay to ask what might seem like a silly question in public. Mm-hmm. Those things I feel like have to be sort of in the air and in the water also. That's so true. And I, I like to say that AI is agile's best use case, uh, because of this. It's like those of us who've been in the agile world and been speaking this language for, for years have been advocating for the psychological safety, for the growth mindset, for the, the experimental mindset.

So these are the conditions in which AI can, can take hold and, and really, uh, thrive. And so it's, it's not unfamiliar territory, and I do feel like things like the AI North Star, um, they're, they're especially important ... today because there's so much fear around AI.

You know, will it take my job? That if you don't put out some, some umbrella statement like, we're gonna... We're using AI to, uh, produce 15 new drugs in five years, the alternative to that is people think we're just going to use it for productivity and to harvest the gains. Mm-hmm. And so you need to create that sense of purpose, the psychological safety, for people to really lean in and want to use this new tool. And we have had some of our clients in the past 12 to 18 months who have tried to go all in on an AI, call it transformation, whatever their- Mm-hmm ... their phrase of the day was, and have found that they could not handle the speed and the, um, forward thinking, all the things that were required to do that because of their operational inefficiencies. They had too much- Mm ... legacy stuff kind of gumming up the works.

And so a few of our clients have had to put those on pause, go back and fix their operational issues- Mm-hmm ... implementing Agile as one component of that, and then feeling like they're ready to go on the next journey- Mm-hmm ... so to speak. So I wonder how much you've seen the... any of these older, bigger, clunkier organizations struggle with some of their legacy ways of working getting in the way of this. Have they had to address that, or can they do it sort of in parallel?

Yeah. It, it's a great question, and I do... You know, this is the five stages of the hyper-adaptive model, and we start off with some, some foundational things like naming your AI leads, but also programmatically supporting them, getting your AI councils in place, but making it dynamic governance instead of status, uh, static governance. And then in stage two, we start to address what you're talking about, which is we start to inject AI into the workflows, and we do that ex- for exactly the reason you're saying, which is the workflows themselves may or may not have ever been documented, visualized, optimized, waste re- removed. And we want people on the front lines to really get intimate with their processes and really own them. And there's a couple of reasons for this. One, in the business process reengineering age, we learned that we can't just bring consultants and reengineer everybody's processes and then leave, because people go back to doing things as they always were. And so getting that frontline ownership is so key. And then I think the other part of that is we know that these processes aren't gonna stand still.

They're gonna be reinvented over and over and over again. So we need the people on the front line to be able to, to reinvent them. And so I think to your point, absolutely, we need to get the processes optimized. You know, people... Another advantage of those who, who know Agile kind of know how to take a look at their processes, know how to work with them, uh, so that we can continually reinvent them for the fore- foreseeable future.

So as we are coming up on the last few minutes of our conversation, I wanna talk a little bit about the good stuff- Mm ... of, of Agile and AI, of the clients that you've been observing and working with, and just what are the good things we can get out of going through this process of change? 'Cause it's always hard. Nobody's really keen to sign up for change.

So I think it's helpful if we can hold out the here's what's on the other side, right? Mm-hmm. And, and here's what's gonna make it all worth it. So from our data's point of view, some of the biggest wins that we are seeing after marketers fully implement AI is when they're on an Agile team, they are over twice as likely to say they can collaborate more easily across teams. So 31% of the Agile respondents told us that versus 15% of the non-Agile respondents. So even as they're bringing AI in, they're not as likely to enjoy those benefits if their ways of working are not fully aligned to achieving these kinds of outcomes.

And same thing with spending more time on strategic work. And to your point about people being kind of fearful sometimes, in an Agile environment at least, we saw 35% of our... The respondents were 35% reporting that they could spend more time on strategic work versus only 14% of the non-Agile respondents. So that's not taking away your job, that's elevating it, which I think is really important.

Uh, and so from your point of view, what other good things are out there, whether you are Agile or not? Like- Mm-hmm ... why should people go through the hard work of doing this change?

Yeah. Well, I think you're, you're alluding to it, which is your job starts to get, uh, easier, and I don't mean in a lazy sort of way, but just in a there's less friction. And I think what your data points to is those folks who've done the hard work of, of Agile, which means we've done the hard work of visualizing the work. We've done the hard work of prioritizing the work.

We know how to manage our capacity. Like, all of that bears fruit in an AI native world. Because when you think about AI native, it's really noisy. AI also just multiplies the number of things you can do. I mean, the, the documents it produces are pages long, and ain't nobody got time for that. And so if you aren't super crisp about what you're doing, how you're doing it, what the priorities are, what you're using AI to do and not to do, then Then it get, it can get really gummy. But if you've got that foundation, all of a sudden it's an, it's an amplifier, right? It's an amplifier for the gunk in your system, it's an ampl- amplifier for all the good in your system, and it's an amplifier for you personally. Like, like, I, I'm gonna pull on the Claude Code Hackathon, and these... this is a global hackathon for people using Claude Code. And the top five winners were a lawyer, a musician, a cardiologist, a civil engineer, and one software developer.

And so that speaks to these capabilities that we all want to do and, and maybe create creative pursuits that we wanna have. And I think that's, that's the Holy Grail, is AI as an amplifier of your dreams. I love that. Amplifier of your dreams. Who doesn't want that?

We all want. Right? Okay. So then last question for you. Uh, I'm sure there are a lot of people who have been nodding along as we're talking and excited about the possibilities, or maybe feeling the drag of an organization that is not maybe on the cutting edge, of- Mm-hmm ... like some of the folks that you describe in your book.

If those folks are just looking for something they can do to get started, maybe their organization hasn't set the big North Star goal yet, but they know that it's important for them and their teams to keep moving and try to stay up to date, what is a kind of, you know, tomorrow when you sit down at your desk, something that those folks might be able to do?

Sure. Um, follow your curiosity, you know, and let that lead, is see, "Hey, I wonder if AI could help me- do XYZ."

Pick that one painful point. Pick that other thing that you really hate to do, and see if AI can help you do it. And of course, third, I'd be remiss if I didn't say pick up the book. You might get a x- a book, Hyperadaptive: Rewiring the Enterprise to Become AI Native, and you might get some inspiration there as well. Absolutely go pick up the book on Amazon, I presume. Yep, or your local retailer, or, uh, anywhere you buy books. Fantastic. And where can people find you, Melissa, if they want to have a follow-up conversation or just follow you for more of these brilliant insights?

Yeah. LinkedIn is always good, and the website is hyperadaptive.solutions. Fantastic. All right.

So as Melissa and I have both pointed out, if you want to keep things moving forward in the right direction, you've gotta know exactly what's going on with your team.

We're talking about visual work management before you can govern, before you can scale. If you're running sprints, dedicate a sprint to this kind of exploration, or at, least set aside some kind of capacity. If you've got champions, maybe consider the kind of governance that Melissa's pointing out is going to be necessary for broader adoption, and effective long-term governance and adoption. A lot of great things to be doing.

Reach out to Melissa if you need help, or you can always find me at agilesherpas.com. And until next time, remember, the struggle is real, but so is Agile, and it's even better with AI.