Issue #5: AI Won't Fix What You Don't Understand

We all love industry buzzwords. A term gets generated, makes its way into corporate decks, and suddenly we’re asked, “So how are we using it?”   

A decade ago, it was digital transformation. Then personalization at scale. Now, it’s hyper-personalization powered by AI. I’m guilty of it too.  In past roles, I championed the importance of personalization to stay competitive. What I didn’t always say out loud? Most brands aren’t ready for it and many don’t even need it.

Personalization and AI (separately and together) have become default marketing ambitions, often disconnected from a clear strategy or a true understanding of your customer.  Everyone thinks AI is this magic bullet, partly egged on by tech providers promising that “AI-powered” anything will solve their problems.

Here’s a real excerpt from a top loyalty tech providers pitch: “Our suite of AI-powered tools empowers you to learn and achieve more…boasts a dedicated AI-focused technology team…integrates AI applications to support loyalty and marketing experts…”. It reads like someone got paid per mention of “AI.”

What those providers rarely say: before you chase AI or deep personalization, you need to get your data house in order. It’s not sexy, but it’s the truth. As Marriott’s Hilary Cook puts it in my interview below, “The barrier isn’t technology. It’s us.”.

No consumer wants communications and offers that don’t appeal to them. So viewing your promotions and marketing through a personalization lens is necessary. But the sophistication varies dramatically by business:

And then there’s what I’m calling ‘over-personalization’: promotions that become so micro-targeted they lose the plot. Starbucks once led the pack in personalized offers, rewarding buying behavior with smart, motivating challenges. But over time, those offers became narrower, shifting from “buy 3 of your favorite drink” to “buy 3 of your favorite drink with a food item you’ve never purchased.”

It’s a pattern across many brands: instead of encouraging customers to buy more of what they already like, we try to steer them toward what we want to move. Ask yourself — are you solving for the customer, or for you?

This problem isn’t a tech issue. It’s a customer clarity issue.   Before adding a whole bunch of fancy tools, ask:

If you answered “no” to any of these, be careful layering on AI. As Hilary says, “AI amplifies what’s already there. If your customer strategy is fuzzy, AI will just make the fuzz louder”.   

Many marketers assume AI will fill the gaps that strategy hasn’t. It won’t. Just like CDP’s were once the cure-all, but marketers didn’t really know how to use them, and so expectations did not meet reality.   Let’s not get over our ski’s again.  There is value in a CDP, and there is huge value in AI. You just have to do the foundational work first to maximize the outcome.

1.   Data Audit:  Distinguish what’s truly reliable (zero- and first-party) from what’s directional (second- and third-party). Label it appropriately.

2. Definitions:  Agree on consistent definitions for core data elements. If one team defines a “new member” as three days old and another as three months, that’s not semantics that’s chaos.

3. Taxonomy and tagging:  Organize and tag data properly so it can be found, reused, and activated.

4. Governance:  Organizational agreement on what “personalization” can and should mean for your brand.

5.  Metrics:  Agreement on measurement and attribution logic.

6.  Collaboration:   Marketing, Analytics, Data, and Tech teams have to work in unison, with everyone bringing expertise together.

Sometimes “personalization at scale” is just marketing trying to justify its tech stack.   But if you ground your personalization strategies in clarity rather than complexity, you’ll see the forest through the trees and get to the capabilities you really need.

Brands I’ve met with that do this best, like Sephora, had to do a lot of unglamorous work to align people, processes, and purpose.    So do the unglamorous first for a big shiny outcome later.

YOUR POV

Read my interview with Hilary below for more on data readiness and the pitfalls to avoid on the path to AI-powered marketing.

And add a comment on whether any of this resonated with your day-to-day role, and how your organization is faring with AI and personalization.

Full disclosure: Hilary used to be my vendor. Back when I was at Dunkin’/Inspire Brands and she was my Client Partner at Epsilon, she managed our loyalty and CRM platforms and operations. She was the one convincing me that taxonomy wasn’t just “back-office weeds,” but the foundation of personalization. We didn’t always agree right away (and that was half the fun), but her perspective always pushed the work forward. Now, at Marriott, Hilary is shaping how one of the world’s largest hospitality companies approaches (true) personalization at scale. You can check out her Adobe Keynote here, which is definitely worth listening to.

#1 SMP: You started out your keynote saying “The biggest barrier (to personalization at scale) is not technology. It’s us”.  Explain this further.

Hilary: What we’re struggling with isn’t new—it’s the work we always knew we needed to do but kept delaying: being clear about what we’re trying to accomplish and why, and ensuring our processes and data support that before we even talk about technology. COVID forced a digital transformation out of necessity, and now AI is pushing us into an even bigger shift. The truth is: outside of your business model and people, success in both digital and AI depends on data. You can’t fake your way through that in the long term. Take a marketing report you rely on—do you know how it’s created, how long it takes, whether the fields are defined consistently, and how confident you are in its accuracy? That report is a product of your business—it’s the visual representation of the health of your processes. Just like in personal health, mobility and consistency may not be flashy, but they’re what prevent repeat injuries and keep you strong for the long run.

#2 SMP: You made taxonomy sound downright thrilling—no small feat!. First, explain what taxonomy is (lest anyone confuses it with taxidermy J) and why it’s so important. And why is getting the basics of process and documentation such a game changer?

Hilary: Speaking of prioritizing your health, this one might be the equivalent of “eat your vegetables.” Taxonomy is simply shared language—the way we define, tag, and organize content, data, and campaigns so they can actually be reused, found, and scaled. Without it, every marketer rebuilds the wheel, and personalization becomes guesswork. With it, suddenly you can automate, measure, and activate across channels. It may feel “in the weeds,” but just like vegetables, taxonomy is the fuel that keeps everything running and makes the flashy stuff possible.

If you want a simple test, pick a term that’s commonly used in your company or grab a few intake briefs—then ask people how they define it. Are you getting different variations? Does it depend on the context? (Bonus points if you ask the same question a week later.) If the answers aren’t consistent, you don’t have a taxonomy that’s permeating your culture. And remember: if intake documents are unclear or inconsistent, you’re putting dirty data into your system from the very start—and no amount of tech can clean that up later.

#3 SMP: You cautioned that AI is an amplifier, not a silver bullet. How do you recommend marketers separate hype from real, scalable use cases?  

Hilary: Admittedly, I can get very passionate about this. AI doesn’t magically solve problems—it amplifies what’s already there. If your data or processes are messy, AI will just make the mess. Many brands talk about using AI to help their workforce—but do you actually know how your workforce does their work? If you’re excited about AI “creating content,” what type of content? (And if you can’t define that, might I suggest the homework from Question 2 😉). Are your brand guidelines clear and consistent, with both good and bad examples? Or are they locked in a PDF that no one can use to train the very models your teams want to experiment with? The best way to cut through hype is to ground AI in real use cases that make marketers’ jobs easier and more effective. Otherwise, you’re chasing buzzwords, not building capability.

#4 SMP: Let’s talk about ROI.  New MarTech is very expensive, so which KPIs matter most to prove the value of deep personalization against the investment? 

Hilary: When it comes to ROI, most people immediately think customer KPIs—conversion, revenue lift, retention. Those matter, but they’re only half the story. The other half is internal ROI: how much faster are campaigns getting to market? How much manual work have we eliminated? How accurate and trusted is the reporting? Sometimes the quickest wins are internal—cutting weeks off launch times or boosting reporting accuracy from 60% to 95%. Those improvements free up marketers to focus on higher-value work, which in turn drives better customer outcomes.

It’s also important to remember that ROI doesn’t live in a vacuum. The objectives you set—and the KPIs that support them—shape behavior. That’s why debating OKRs (Objective + Key Results) is critical. There’s a real difference between an objective and its supporting metrics, and common language here is vital. Just as important is pressure-testing both the intended and unintended incentives of any KPI set. Because what you measure is what people will optimize toward—whether it was the outcome you wanted or not.

#5 SMP: For smaller brands that don’t have the same people and financial resources as Marriott, how can they still approach your “pilot, scale, business as usual” model to reach maturity? And with tech moving so fast, how should they think about investments that could feel aged by the time they implement it?

Hilary: You don’t need a massive budget or team to build maturity—the pilot → scale → business as usual model works no matter your size. The key is focus. Pick one audience, one channel, and one use case, then prove it out. Once you’ve got results, scale to the next. The temptation is to chase shiny new tech, but remember: every new tool is only as good as the data, content, and processes you put into it. With tech moving fast, the safest investments are in capabilities that compound and are designed to evolve. Flexibility is everything. And if your Technology and Digital teams sit separately from Marketing, they need to learn to build with marketers, not for them.

It’s a lot like health—you don’t need a luxury gym membership to get fit. Walking consistently, eating vegetables, and stretching will get you further than any expensive quick fix. Smaller brands may actually benefit from their size and perceived constraints: fewer silos, less legacy complexity, and faster decision-making can all make it easier to stay consistent and scale what works.

SMP Bonus Question: You’ve been both on the brand side now with Marriott and the partner supplier side when you were at Epsilon and elsewhere. What’s one thing you wish marketers knew about their partners, and vice versa?

Hilary: An executive recruiter once told me that brands assume agency folks don’t know how actually to deliver, and agencies assume clients leave at 4pm and don’t keep up with the trends. From personal experience, both are the worst ends of tired stereotypes. What I do wish is that consultants and agency partners asked more questions to understand business context before presenting recommendations. And on the flip side, I wish brand teams treated partners as true extensions of their team—sharing what they’re solving for instead of keeping it at arm’s length. If your first reaction is “but they’ll use that to upsell me or take advantage of me,” then you don’t have the right partner. The best work happens when there’s mutual trust and a willingness to build with each other.