Please Pick Your Lane: AI, Martech, and the Future of Enterprise Software
•Rachel Michaela Gray•Season 3•Episode 5
Use Left/Right to seek, Home/End to jump to start or end. Hold shift to jump forward or backward.
0:00
|
20:29
AI isn’t slowing down, but a lot of enterprise decision-making is. I’m seeing marketing leaders, CIOs, CMOs, architects, and security teams collide around the same problem: we want AI use cases now, yet our governance and implementation choices are turning into long delays. When an AI vetting process stretches from four to six weeks into four to six months, the business doesn’t just lose time, it loses learning speed, campaign momentum, and confidence.
I dig into a more practical approach to enterprise AI governance by separating use cases by risk and autonomy. AI-assisted content creation with non-sensitive inputs should not face the same gates as AI-powered marketing intelligence pulling from approved reports, and neither should be reviewed like agentic campaign execution where an AI system can take actions across the enterprise. The point isn’t to relax security, it’s to make data security, permissions, monitoring, and data-flow mapping match the actual stakes, with InfoSec involved during use case development instead of after the fact.
From there, we tackle the reality of the fragmented enterprise. Multiple brands, regions, tech stacks, and reporting structures create silos where breakthroughs get lost in email and duplicated by other teams months later. A shared, authorized organizational knowledge layer can help marketers find what already worked, accelerate experimentation, and reduce duplicate work even when systems don’t perfectly integrate. We also get honest about build vs buy for AI: just because a team can build something doesn’t mean it should, especially when maintenance, infrastructure, upgrades, and support turn a demo into a long-term burden.
I close with what this means for the future of enterprise software and SaaS platforms: flexibility and ecosystem readiness matter more than ever as AI stitches tools into continuous workflows. If this resonates, subscribe, share the show with a teammate, and leave a review so more marketing and technology leaders can find the conversation.
Hello and welcome back to season three of That Martech Girl, the podcast where Martech Meets Culture, the business podcast with personality. My name is Rachel Gray, and today we are going to be diving into another solo episode. In the first one, we were talking about how everybody wants an AI strategy. How exactly is that practically accomplished? Today we're going to be moving beyond the strategy conversation because I've been talking to a lot of marketing leaders, engineering teams, CIOs, CMOs, architects, security stakeholders. And what I'm seeing is this tension between the AI use cases and the implementation decisions. Because everybody wants AI. Everybody has a specific use case they think is super important. But there is this argument between what it should look like, who owns it, and whether or not to build or buy the technology, right? And so as AI becomes increasingly capable, the decisions don't get easier. They actually get a lot harder. So today I want to talk about four things. The first thing I want to talk about is governance. The second thing I want to talk about is in an organization where you have multiple business units, how exactly can you use AI to bridge the gap between those business units? And then thirdly, I want to talk about the build versus buy conversation. And then finally, I want to talk about the future of enterprise software and how flexibility is most important. So
let's start with something I think is absolutely necessary and that we need to approach more intelligently, and that is governance. I truly appreciate rigorous enterprise AI vetting because I understand why you're doing it. You're protecting your data, you're protecting your company, and you are trying to set a standard for how things will be governed in the future. And also, the news is scary when it comes to AI. However, over the last couple months, I have noticed that these AI vetting committees and processes are taking longer than before. What I have seen typically is maybe a four to six week process, which is already long. But now I am seeing four to six months, which, if not handled correctly, it can turn into a year. And when we think about the pace and acceleration of the world we're living in as relates to AI, that is simply unacceptable for everybody involved, for the organization and for the people who are spending their hard earned time and energy in these processes. And I think the reason this is happening is because every single AI use case is being treated in the same way and being vetted the same way. And I don't think that is the best approach. Let's let's talk about this and break it down into some common scenarios and use cases, And I will show you just like how different they are and how the vetting needs to happen differently based on the use case. So
one typical use case I see is like AI-assisted content creation, right? So using AI to brainstorm ideas, concepts, drafting email variations, and using non-sensitive information, right? You're not really getting any proprietary data access, none of that. But there are concerns, right? There are brand standards that are needed, there's accuracy needed, IP, appropriate data handling, but none of this is extremely sensitive. Then the next use case I want us to think about is AI-powered marketing intelligence. Now, this is a bit more involved where AI is accessing approved reports, historical campaign information, experimentation learnings, and marketers are getting questions and recommendations from organizational knowledge, right? So the concerns here would be more around data permissions, reliability, and then separation of this information across different business units. Now, the last use case and the most involved one I want us to think about is agentic campaign execution. So this is now creating an agent that can help create assets, update content, and it's pretty much working autonomously throughout your entire enterprise. Right. And so the stakes are going to change here because this AI agent can execute actions rather than having a human ask it questions and it give recommendations. However, what I have seen, and I've been a part of processes for each of these use cases, the governance seems to be the exact same. And the AI committee process for approval seems to be the exact same. However, I don't think it should be because there are different autonomy and risks associated with each one's. I think it's really important that security teams really understand what each use case and capability does so that they access and assess what exactly the risk is in each use case. And then, secondly, business teams need to be able to accurately explain the intended use case and the business outcome. Because if not, then can you really blame the security teams? I don't think so. And then, of course, architects and engineers need to map the data flows, the system access, and permitted actions for every single use case. But all of this needs to happen not in a silo when the use cases are being created and developed. It needs to happen at the same time, right? I find that InfoSec processes like happen after the fact. I think it needs to happen during the use case development. So now speaking of architecture, there's another part of this conversation that I think is really important.
And that is how to connect a fragmented enterprise with AI. So let's talk about that next. One of the biggest hurdles that I hear on a day-to-day basis from the marketers I talk to is that they're working in these organizations that are extremely fragmented. And this is normal. I hear this all the time. The reason being when you have a large multinational organization, you typically have multiple brands, regional marketing teams, different tech stacks, and different reporting structures, especially. So this creates silos and a very fragmented way of working. So I have an example of this, and it has to do with say one brand in this large organization discovers a new experimentation breakthrough, right? And it they're seeing great conversion rates because of this. Now, how does another brand learn about this? Typically, this is going to happen in an extremely manual manual way, right? Someone's going to document the experiment and communicate results. Then they're going to interpret the findings and say, oh, maybe this would apply to another business unit. And then they're going to maybe send via email their findings. It's probably going to get lost in email. And typically, this process is going to take months to get communicated to another business unit. And so historically, you have business units maybe creating the same experiments and trying to get to the same results. And so there's all of this work being done that's duplicate and doesn't need to be done. Now, in a world of AI, what I'm seeing happen that I think is the biggest help to these large enterprises is this shared organizational knowledge that can be transferred very, very easily. And I would call it authorized organizational knowledge because this enables marketers in every single business unit to now have this repository of information to call from and to say things like, hey, what have other brands in our organization learned about improving acquisition? Things like, which experiments succeeded with similar customer segments? Right. And so asking these questions in a way that pulls from the shared organizational knowledge, it really, really complements the large enterprise and causes learnings to accelerate throughout these types of organizations. Now, I'm not saying that AI is going to magically integrate in compatible systems because if everybody has a different tech stack, I'm sorry, that's sometimes just like not going to work. However, it can make knowledge sharing so much easier. And it can eliminate the duplicate work that's usually happening in these fragmented enterprises. So instead of relying on these manual processes where you have all these meetings, documentation, and traditional knowledge management, you can now call on this shared repository. But even when organizations are seeing this opportunity, we're running into another challenge, and everybody suddenly wants to just build AI from the ground up.
And I don't think I necessarily agree with it. So let's get into it. I'm very passionate about this topic because it's something that I'm hearing on a day-to-day basis. I work with such smart people, typically engineers who can do absolutely anything. I am dead serious. Like I think that they could build Chat GPT if they wanted to. Absolutely, they could. They absolutely could. But does that mean that they should? And that is that is the the main issue that I want to talk about today. Just because you can, should you? Just because you can build AI from the ground up, should that be what you are doing? Right? Because I think typically, you know, engineers they love solving difficult problems. And I am a scientist at heart, so I understand and I resonate and I have empathy for that. However, a working demo is not the finish line. You also need a lot of other things. You need security, you need monitoring, you need infrastructure, you need upgrades, you need support, you need maintenance. And the time that you're taking to build something that is not in your business model, you could be taking that time to work on something that is actually giving your business true ROI. So before you even start a project, my question always is: are we doing this because it gives us an advantage? Or are we doing it because we're super excited and because we're smart enough to do it? We were just talking about the large enterprise, and this is a problem I see happening all the time. And it has to do with competing priorities, where one team will say, Yes, we have identified a business problem and we have identified available AI capabilities. Let's go in that direction. Then another team will come and say, Actually, guys, we are capable and we can do an internal build and do the same thing, but with our internal resources. Now the whole organization has these competing priorities. One group wants to move forward, another one wants to experiment, engineering wants to evaluate what it can build. And then, of course, leadership has to decide what approach makes sense. And in the meantime, the business problem is forgotten, and the execution that was thought of in a specific timeline is extremely slowed down. So my question is: at what point does experimentation stop being innovation and it start becoming a delay? Who decides when to build, what to build? What happens to the business units that are ready to move on this opportunity? And are we measuring this cost of indecision? It's really frustrating to see innovation stall because organizations are more interested in playing around in the sandbox and experimenting with passion projects than actually moving forward and solving the business problem. Now I'm not saying don't build anything because I think especially with AI, it's very malleable. And you need to be able to customize it to your specific unique operational models. Like if you're not doing that, then you're not going to get an output that serves your business. So you do need to, of course, modify what is existing and out there. But what I'm saying is don't recreate every underlying technology capability because that is truly a waste of time. And if that is not in your business model, then you are wasting time and energy and resources. However, I'm not saying never build because sometimes internal development is really justified when there's no existing solution that meets your requirements, when maybe the capability is defensible, differentiated, and when in some cases long-term economics are favoring the internal build, then that is a situation where the decision is to build. But typically, there are AI companies that have already solved these problems. And it is up to you to then take that capability and uniquely craft it to your business problem and business need. Now I spoke very, very passionately to the enterprises. I think it's time to speak passionately as well to the SaaS providers, of which I am a part of. So
I think that in today's world, there is a new expectation. This is a completely new ball game. And now more than ever, flexibility is extremely important for us as SaaS providers. Because traditionally, organizations are buying a CMS, they're buying a marketing automation platform, they're buying a CDP, they're buying an experimentation tool, and maybe an analytics system, right? And each of these tools typically had their own interfaces, workflows, data structures, integrations, et cetera. But now with AI, you can stitch all of this together into one continuous, autonomous workflow. Right? So the planning, the creation, the experimentation, the measurement should all be happening, of course, with humans in the loop, but in an autonomous way that connects all of these pieces together. So if you are one piece of this puzzle, it is so essential now more than ever before to play well in the ecosystem because vendors are asking questions like, can this platform connect with the rest of our ecosystem? Things like, does these capabilities participate in broader AI-powered workflows? And can you really answer that? I think, you know, before we were saying, um, how do we bring more people into our ecosystem? How do we have them buy more of our products? I think it's really about flexibility now because of this new age. I think the future of enterprise software isn't just what your platform can do, it is what enables the larger organization to accomplish. Today
we spoke about governance, we spoke about enterprise intelligence, we also spoke about the build versus buy conversation. And finally, we ended on creating a flexible ecosystem in today's new AI-centered world. And I think the common theme here is that all of these topics are not just IT decisions anymore. They truly affect marketers' ability to work, to personalize, to run experiments, to execute campaigns, and ultimately to really deliver on the customer experience that is the art of why you are a marketer in the first place. I don't think that marketers need to be CIOs in any way, shape, or form, but increasingly in today's world, understanding the role of the CIO and how exactly to craft your use cases in ways that speak to the governance that is needed for the organization is more important now than ever before. And on the other hand, I think it's important for CIOs to understand the operational realities and needs of the marketing organization. And I think this is a great segue to our next episode, which we are interviewing Brandon Scott. He's already been on the podcast last season. He is coming back. He is a CIO and also a CGO as well. He works for 10 Adams. Um, he's been a great technology leader and he's been operating at the intersection of enterprise strategy, architecture, business growth, and innovation. So I think he's a great person to continue this conversation. I'm very excited about next week's episode. I hope you are too. But thank you so much for listening to this episode of That Martech Girl. I really, really appreciate your support. Um, continue to follow us on social media as well as all of the different podcast platforms. I truly appreciate your support. Have a wonderful rest of your week. Thank you. Bye bye.
