Build for States, Not Segments: The Customer Decisioning Architecture Startups Get Right (And Enterprises Get Wrong)

You start simple. "Prospects" and "Customers." Two segments. Clean. Manageable.

Then someone asks: "Should we treat engaged customers differently than lapsed ones?" Sure, makes sense. Now you've got four segments.

Then: "What about high-value versus low-value?" Okay, eight segments.

Then product interests. Geographic nuances. Lifecycle stages. Channel preferences. Before you know it, you're maintaining 200+ segments, your campaign calendar looks like a Gantt chart from hell, and your creative team is drowning in requests for "just one more email variant for the West Coast high-value SUV owners who opened the last email but didn't click."

Sound familiar?

Here's what nobody tells you when you're starting out: Enterprise companies hit this wall at around 100,000 customers and spend millions trying to fix it. They hire consultancies. They buy new platforms. They reorganize their marketing ops teams. And they still can't escape the segment trap.

Startups have a chance to avoid this entirely—but only if you build differently from day one.

The alternative isn't "better segmentation." It's a fundamentally different mental model: state-based decisioning. And if you're a founder or early-stage operator, you have a narrow window to get this right before you scale into the same mess that's strangling enterprise marketing teams.

Let me explain what I mean.

---

What Segments Get Wrong (And Why You Inherited This Model Anyway)

Segmentation isn't stupid. It's just outdated.

Segments made perfect sense in the direct mail era. You literally had to physically sort mail into bins: "Prospects go here, customers go there, high-value customers in this special bin." Digital marketing inherited this mental model without questioning whether it still made sense. And every major marketing platform—Salesforce, Adobe, HubSpot, Braze—was architected around it.

So when you're setting up your first marketing automation system, segments feel like "the way." The platform UI nudges you toward them. Your agency recommends them. That's just how it's done.

But here's the problem: the constraints that made segments necessary are gone. You're not sorting physical mail anymore. You're not limited by batch processing. You have real-time data, event streams, and the ability to make decisions in the moment.

Yet we're still using a framework designed for 1990s logistics.

What Segments Actually Are

Let's be clear about what we're dealing with:

• Static buckets based on attributes or behaviors
• Require constant maintenance as you add dimensions
• Force you to choose when someone fits multiple segments (the dreaded "primary segment" problem)
• Scale poorly through combinatorial explosion
• Miss the context of where someone is right now

Quick terminology note: When I say "state-based decisioning," I mean decisioning based on relationship states—like "at risk" or "service due." I'm not talking about geographic states, though honestly, "lives in Chicago in February" probably should trigger warm-weather vacation offers.

Here's a real example from my world in automotive marketing:

Someone bought a car two years ago. In a segment-based system, they land in: "Customer - 2 Year Owner - Midwest - SUV."

Okay, great. Now what?

Are they:

• Happy and just need a routine service reminder?
• Unhappy and showing churn risk?
• Starting to research their next vehicle?
• Completely satisfied and prefer minimal contact?

The segment tells you almost nothing about what to do next.

The Yearbook Photo Problem

Here's how I think about it: Segments are like yearbook photos. They capture who someone was at a moment in time. Useful for recognition, terrible for action.

States are like your Ring doorbell. They show who's at your door right now and what they need.

When someone rings your doorbell at 11pm, you don't pull out their yearbook photo. You look at the camera and respond to their current situation.

That's the difference.

The N-Squared Problem

The other issue with segments: they multiply in ways that get out of control fast.

• 5 segments = manageable
• 10 segments = 100 possible combinations to think about
• 20 segments = 400 combinations
• 50 segments = you're drowning

You know you've hit peak segment complexity when you need a Venn diagram with 17 circles just to figure out which email someone should receive. At that point you're not doing marketing—you're doing set theory homework.

---

State-Based Decisioning: The Alternative Mental Model

Okay, so segments have problems. What's the alternative?

Here's the core idea: Segments are like a census—they tell you who's out there. States are like traffic lights—they tell you what's happening RIGHT NOW and what you should do about it.

You wouldn't navigate Chicago using census data from 2020. You'd check the traffic. But most marketing systems are still trying to drive using demographic snapshots instead of real-time signals.

What Is a "State"?

A state describes where someone is in their relationship with you right now. Not their demographics. Not their historical behavior category. Their current context.

Key characteristics:

• Contextual, not demographic
• Can overlap (someone can be "Service Due" AND "At Risk" simultaneously)
• Have priority hierarchies (red light beats yellow beats green)
• Designed around decision-making, not reporting

Let me show you the difference with a side-by-side comparison.

Segments vs. States: An Automotive Example

SEGMENTS APPROACH:

• Prospect - Shopper
• Prospect - In-Market
• Customer - New (0-6 months)
• Customer - Established (6-24 months)
• Customer - Loyal (24+ months)
• Customer - Service Due
• Customer - At Risk

And this is the simple version. In reality, you'd multiply these by product type (sedan, SUV, truck), geography (regional offers), value tier (economy, premium, luxury), channel preference (email, SMS, app)...

You see where this goes.

STATES APPROACH:

• Exploration (researching, not ready to buy)
• Evaluation (comparing options, high intent)
• Transaction (in active purchase process)
• Onboarding (first 90 days of ownership)
• Engaged Owner (active, satisfied)
• Service Need (routine maintenance due)
• Concern (expressed problem or complaint)
• Risk (showing defection signals)
• Advocacy (referring others, promoting brand)

Notice What's Different

These states describe the relationship condition, not the customer profile. They're fewer in number but contextually richer. They can overlap. And they're explicitly designed around the question: "What should we do next?"

Let me walk through an example using the traffic light metaphor:

Someone bought a car two years ago.

The segment view: "Customer - 2 Year Owner - Midwest - SUV"
That's demographic data. It's like knowing their address and what's in their driveway.

The state view might show:

• Service Due (yellow light - proactive reminder needed)
• Recent Complaint (red light - retention risk)
• High Engagement (green light - advocacy opportunity)

Now you know what to do. The red light beats the yellow light beats the green light. Address the complaint first, then offer service scheduling, then (maybe) ask for a referral.

You're making contextual decisions, not executing pre-planned campaigns.

Priority Hierarchies Are Critical

Just like traffic signals have rules (red always overrides green), states need priority logic:

• Concern/Issue states override everything (don't upsell someone who just filed a complaint)
• Transaction states beat routine communication (don't interrupt a purchase flow with a survey)
• Advocacy opportunities come after needs are met (earn the right to ask for referrals)

This isn't arbitrary—it's customer-centric design baked into your architecture.

---

Why This Matters for AI (And Why I Keep Hammering on This)

This is where state-based thinking transforms from "interesting architecture idea" to "competitive advantage."

If you've read my other posts, you know I'm all about predictive AI as the workhorse while everyone else chases generative AI demos. State-based decisioning is how you actually put that predictive power to work.

Predictive AI Works Better With States

Compare these two questions:

Segment-based: "Predict who will click this email campaign"
State-based: "Predict who is transitioning from Engaged Owner to At Risk"

One optimizes a campaign. The other prevents churn. Different game entirely.

Why does this matter?

• ML models are predicting transitions, not just responses
• You're asking about relationship dynamics, not campaign performance
• The predictions are more actionable (you can intervene on state changes)
• You're building toward customer lifetime value, not just next conversion

Content Personalization Becomes Contextual

With segments, you end up building 47 email variants for different segment combinations. Your creative team drowns. Testing becomes impossible because you've got too many variants and not enough volume in each.

With states, you build modular content blocks that assemble based on current state. Einstein Personalization (or equivalent) picks the right module. Way less creative burden. Way more relevant to the recipient.

Remember my post about moving from "If/Then" to intelligent predictions? State-based architecture is how you actually enable that. You're not writing brittle rules for every segment combination—you're letting the AI pick content based on current context.

Orchestration Becomes Smarter

Segments force calendar-based campaigns:

• "Run Q2 reactivation campaign to lapsed segment"
• Everyone gets it on the same day regardless of their situation

States enable trigger-based orchestration:

• "When someone enters 'Risk' state, initiate retention sequence"
• "When someone is in both 'Service Due' and 'Engaged Owner,' send convenience-focused reminder"

Your email templates get simpler because decisioning happens in the data layer, not the template layer. Less AMPscript spaghetti. More maintainable. Less brittle.

It's like moving from a spreadsheet full of nested IF statements to a proper database with views. Same information, way better structure.

---

Why Enterprises Can't (Or Won't) Build This

Now here's the painful part: if state-based decisioning is so much better, why isn't everyone doing it?

The reason isn't technical. It's political.

The IT Problem

Legacy architecture is the first barrier. Existing marketing systems—Salesforce Marketing Cloud, Adobe Campaign, all the major platforms—are architected around segments and campaigns. The data models, the UI, the reporting, everything assumes this structure.

Rebuilding that is a massive replatforming project. And in enterprise IT, "massive replatforming project" translates to "career-threatening risk."

Nobody wants to own it:

• IT doesn't want to own a marketing data model redesign
• Marketing doesn't have the technical chops to spec it properly
• The data team is already underwater with other requests
• It falls between org silos, so it goes nowhere

The result: "Why would we tear down what's working?" (Even if "working" means "barely functional and getting worse every quarter.")

In enterprise, suggesting a fundamental architecture change is like suggesting your extended family switch Thanksgiving to a different house. Technically possible. Logically sound. Never going to happen.

The Agency ATM (This Is the Part Nobody Talks About)

But here's the really inconvenient truth: Your agency has zero incentive to recommend this.

Their business model depends on complexity and volume.

Think about what they bill for:

• Creative concepting for each segment variation
• Copywriting for 47 different email versions
• Design and production for all those assets
• Campaign deployment and QA
• Performance reporting on all those segments
• "Strategic segmentation workshops" (annual revenue stream)

State-based decisioning with modular content? That could reduce their ongoing creative requests by 60%+. That's 60% less revenue.

So what do they recommend instead?

• More personas (more concepting workshops)
• More segmentation refinement (more strategy retainers)
• More journey mapping (more billable discovery)
• More campaign "innovation" (more production work)
• More creative testing (more variants to build)

All of which keeps the ATM humming.

Your agency's strategic recommendation always sounds sophisticated: "We should develop micro-segments based on psychographic clustering and propensity modeling."

Translation: "We'd like to bill you for six more months of creative production."

Look, agencies aren't evil. They're rational economic actors. But expecting them to recommend an architecture that kills their most profitable revenue stream is like expecting a lawyer to recommend you stop getting divorced. Technically good advice. Bad for business.

The Frozen Middle

Meanwhile, inside the enterprise:

• Marketing leadership doesn't understand the technical difference
• Operations teams are too busy keeping the current system running
• Nobody has bandwidth for "innovation projects"
• Career risk looms: if you push this and it fails, you own it
• Easier to complain about complexity than fix the foundation

A weird alliance emerges:

• Agencies don't want it (kills revenue)
• IT doesn't want it (too hard, too risky)
• Creative teams might not want it (less variety, less visibility)
• Executives don't understand it, so they fund what agencies recommend

The only people who actually want state-based decisioning:

• Operations people drowning in campaign management
• Lifecycle marketers frustrated by context-blind communications
• Data people who see the scalability cliff coming
• Customers getting the wrong messages at the wrong times (but they don't get a vote)

So you've got IT saying "too hard," agencies saying "not necessary," and executives hearing "we need more sophisticated segmentation."

Meanwhile, operations is drowning, customers are getting service reminders mixed with conquest offers, and nobody's measuring the cost of this complexity.

Enterprises are trapped. They know segmentation is breaking. But the organizational antibodies prevent the cure.

---

The Startup Advantage (And How to Actually Use It)

But if you're a startup? You don't have these antibodies yet.

Here's your unfair advantage: You get to build this right the first time.

Your Structural Advantages

1. No Legacy Architecture

You're choosing a marketing platform now. You can design for states from the beginning. No migration pain. No political battles to refactor existing systems. Clean slate is worth more than you think.

2. No Agency Dependency

You're building lean, modular content anyway—you have to, because you don't have budget for 47 variants. Your constraints force better architecture. Efficiency isn't a threat to your business model. It is your business model.

3. Smaller Data Sets (For Now)

It's way easier to model states with 10,000 customers than 10 million. You can test fast. Iterate quickly. Mistakes are recoverable. You can even talk directly to customers to validate your state definitions (try doing that at Ford).

4. Direct Control

No IT approval process. No change management theater. No cross-functional alignment meetings with 17 stakeholders. You decide. You build. You ship.

You have a chance to install the traffic light system from the beginning. Enterprises are stuck trying to retrofit traffic lights into a census database while keeping 10 lanes of traffic moving.

It's way easier to install the lights before you pave the roads.

How to Actually Do This

Okay, enough theory. Here's the practical playbook.

Step 1: Start With State Definitions, Not Segments

Map the actual relationship states in your business. Keep it simple—5 to 8 states max to start. Focus on the decision-making question: "If someone is in THIS state, what should we do differently?"

Define priority hierarchies: which states override others when someone is in multiple states simultaneously.

Document state transition triggers clearly.

Example for a SaaS product:

• Trial (first 14 days, exploring product)
• Activated (completed key onboarding action)
• Engaged (regular usage pattern established)
• Stalled (no usage in past 7 days)
• At Risk (usage declining, or unresolved support issue)
• Expansion Opportunity (hitting plan limits, or using advanced features)
• Churned (cancelled subscription)

For each state, ask: What should our communication or action be?

Step 2: Choose Tools That Support This

Most modern platforms can be architected this way—Customer.io, Iterable, Braze, even parts of HubSpot. But their UX often nudges you toward segments because that's the inherited mental model.

Look for:

• Event-driven capabilities (not just batch/campaign execution)
• API-first architecture (so you can build custom logic outside the platform UI)
• Flexible data models (not rigid segment hierarchies)

Warning: You'll have to be intentional. The platform will want you to build segments. Resist.

Step 3: Build Modular Content From Day One

Think content blocks that assemble, not campaign-specific creative assets.

• Header + Body + CTA, where each component can vary independently
• Design system that supports this componentization
• Make it easy to test and iterate on individual components

Resist the temptation to build custom campaigns for every scenario. Build the Lego blocks first. Assemble them based on state.

Step 4: Instrument State Transitions

Track when people enter and exit states as explicit events in your data layer.

Measure state progression: Are people moving toward Engaged? Toward Expansion Opportunity? Toward Advocacy?

Optimize for state health, not campaign metrics.

"What percentage of Trial users reach Activated within 3 days?" is way more useful than "What percentage clicked the email?"

Step 5: Keep It Simple

Don't over-engineer. Ship and learn. Add complexity only when the data shows you need it.

You're building a foundation, not a finished cathedral.

Startups have a huge advantage here: nobody's built a segmentation machine yet that needs to be taken out back and shot. You can just... not build the wrong thing.

---

The Competitive Moat This Creates

Here's what you get if you do this right:

When you scale, you scale cleanly. Going from 10,000 customers to 100,000 doesn't break your system. You're adding complexity linearly, not exponentially. Your operational costs stay manageable. You don't hit the wall enterprises hit.

Your customer experience is better. Contextually relevant communications. You're not sending service reminders to someone who just filed a complaint. Less noise, more signal. Better engagement. Better retention.

You're positioned for AI leverage. When you're ready to add predictive models, you already have clean state-based data. State transitions are perfect training data for ML. You're building toward sophistication, not trying to escape chaos.

You look professional to enterprise buyers. If you're building MarTech SaaS, this architecture signals maturity. Enterprise buyers recognize state-based thinking—even if they can't build it themselves. "We support state-based decisioning workflows" becomes a genuine competitive differentiator.

Meanwhile, your enterprise competitors are stuck in segment hell, paying agencies millions to manage the complexity, fighting IT battles to modernize legacy systems, and drowning in ad hoc campaign requests because everything's too brittle to automate.

---

The Boring Work That Wins

State-based decisioning isn't sexy.

It won't get you on stage at SaaStr. Your creative team won't put it in their portfolio. There's no viral LinkedIn post about your innovative segmentation methodology. No case study with impressive-sounding frameworks and acronyms.

What it is: infrastructure.

It's the plumbing of customer engagement. The foundation that lets everything else work. The boring work that compounds over time.

If you've read my other posts, you know I think about this stuff through a Chicago lens. Predictive AI is the "Chicago of AI"—not flashy like Hollywood, but doing the heavy lifting that makes everything else possible. State-based architecture is the same energy.

It's not Hollywood-flashy. It's big shoulders. It's the work that makes everything else work.

The companies that will dominate the next decade of marketing aren't the ones with the flashiest creative or the most sophisticated segmentation taxonomies. They're the ones who built the right foundation early—and didn't have to rebuild it later.

You have a window to be one of those companies.

Don't waste it building the same broken architecture that enterprises are desperately trying to escape.

---

Segments gave us maps. States give us GPS.

Both show you where customers are—but only one helps you decide what to do about it in real-time.

And just like with GPS: once you've used it, you'd never go back to paper maps.

---

What are you building? I'd love to hear how you're thinking about customer decisioning architecture. Comment below or you can find me on LinkedIn or at mikehotz.com.

Why Trust Will Decide the Future of AI in Automotive Marketing

Artificial Intelligence is no longer just an experimental tool in the auto industry—it’s already in the showroom, on dealership websites, and woven into marketing campaigns from the world’s biggest OEMs. From predictive analytics that optimize inventory to chatbots guiding shoppers through vehicle research, AI promises efficiency, personalization, and scale.

But there’s one obstacle automakers can’t afford to ignore: trust.

---

The State of Public Confidence in AI

Recent research paints a clear picture—Americans are skeptical.

• Only 17% of U.S. adults believe AI will have a positive impact on society in the next two decades (Pew, 2025).

• Just 2–3% of Americans strongly trust businesses or government to use AI responsibly (Gallup, 2025).

• Nearly 80% want AI safety and data protections prioritized over rapid development.

That skepticism shows up every time a consumer interacts with an AI-driven website, app, or dealership chatbot. Convenience may win clicks, but privacy concerns and a lack of transparency quickly erode confidence.

---

What This Means for Car Buying

The car-buying journey is already stressful and high-stakes, and research shows AI adoption follows a clear pattern:

• Shoppers are open to AI early in the process—research, comparisons, browsing inventory.

• Trust drops sharply during the “Buy” stage, when financial decisions like pricing, financing, and trade-ins are on the table (CarEdge, 2025).

• While 83% of consumers expect AI to impact car buying within the next decade, only 37% of dealers see it as central to their operations (Cox Automotive, 2025).

That mismatch suggests a widening gap: consumers want AI-enabled experiences, but only if the trust equation is solved.

---

Predictive, Generative, and Agentic AI: The Toolkit

The industry is experimenting across three main types of AI:

• Predictive AI works in the background, forecasting demand, optimizing pricing, and identifying the right offer for the right buyer.

• Generative AI personalizes communications—emails, chat interactions, even walkaround videos. It feels magical, but disclosure is often minimal.

• Agentic AI acts autonomously across channels, serving as a 24/7 “virtual showroom concierge.” Customers notice the benefit (instant help), not the mechanism.

Most OEMs and dealers emphasize the outcome—faster, more relevant experiences—rather than the process. That works, but only up to the point where transparency becomes necessary.

---

Building Trust: Where OEMs Need to Focus

Automotive brands can’t treat transparency and privacy as afterthoughts. Trust is the differentiator.

1. Disclose strategically.
   When AI influences major financial decisions, disclosure isn’t optional—it’s critical. A cautionary tale comes from outside the industry: Zillow’s failed home-buying program. The company relied on AI to predict home values but failed to disclose its limitations. Consumers and even Zillow itself trusted the outputs blindly. When the model proved inaccurate, the program collapsed—erasing billions in value and damaging trust. For automakers, the lesson is clear: don’t hide the AI when it touches pricing, financing, or trade-ins. Transparency at these high-stakes moments preserves credibility.

2. Offer human fallback.
   A seamless handoff to a human advisor in the “Buy” phase reassures customers that they remain in control.

3. Adopt privacy-first practices.
   Techniques like federated learning and differential privacy allow personalization without compromising sensitive data.

4. Make trust a brand asset.
   With 67% of Americans disabling cookies, transparency itself can become a selling point.

---

Final Word

AI in automotive marketing is here to stay. But adoption will stall if trust doesn’t keep pace with innovation. The OEMs and dealers who strike the right balance—delivering personalization with transparency, privacy, and human touch—won’t just sell more cars. They’ll earn something far more valuable: lasting consumer confidence.

---

What’s your take? Would you trust an AI-driven car-buying journey if the dealer offered full transparency and a human fallback option?

From "If/Then" to Intelligent Predictions: The AI Revolution in Email Personalization

 For years, email personalization has been the holy grail of digital marketing. We've all strived to send the right message to the right person at the right time. But how we achieve that "rightness" is undergoing a massive transformation. We're moving beyond the laborious, often brittle world of "if/then" rules and embracing the power of predictive AI.

At my core, I'm all about making marketing smarter and more effective. And the shift to AI-driven email personalization is one of the most exciting developments I've seen in years.

The Era of the "If/Then" Labyrinth

Think about how we've traditionally approached email personalization. It's been a massive, intricate flowchart.

• "IF a customer bought Product A, THEN recommend Product B, UNLESS they also bought Product C, AND EXCLUDE anyone who visited the returns page in the last 7 days."

Sound familiar? This "if/then" approach, while foundational, comes with significant baggage:

1. Scalability Headaches: As our customer base grows and their behaviors diversify, the number of rules explodes. Managing this becomes a full-time job, leading to inevitable errors and missed opportunities.

2. Rigid & Reactive: These rules are static. They can only react to predefined conditions. They can't anticipate a customer's evolving needs or adapt to subtle shifts in behavior in real-time.

3. Limited Personalization: We're often personalizing to broad segments, not individuals. A rule might apply to thousands, but does it truly resonate with each one? The depth of personalization is inherently limited.

4. Maintenance Nightmares: Updating, debugging, and ensuring these complex rule sets don't conflict is time-consuming and prone to breaking existing logic. Exclusions and suppressions add layers of administrative burden.

5. Underutilized Data: We collect vast amounts of data, but with "if/then" rules, we only leverage the data points that fit our predefined categories. Nuance and subtle signals are often lost.

This creates a self-perpetuating cycle of complexity, where our time is spent managing rules rather than innovating.

Enter Predictive AI: Your Intelligent Email Navigator

Now, imagine an email system that learns from all your customer data – not just what you explicitly tell it to look for, but also the hidden patterns and correlations. This is the magic of predictive AI. Instead of us dictating every step, the AI acts as an intelligent navigator, predicting the most relevant and impactful content for each individual.

Here's how this intelligent approach transforms email personalization:

1. True 1:1 Personalization at Scale: AI moves us from segment-level personalization to individual-level personalization, creating highly relevant experiences for millions of customers simultaneously, without manual effort for each new permutation.

2. Proactive & Adaptive: AI anticipates customer needs and adapts to changing behaviors. If a customer's interests shift, the AI recognizes this and adjusts its recommendations accordingly, far faster than we could update manual rules.

3. Unlocking Data's Full Potential: The AI can process and learn from massive, complex datasets, identifying subtle signals and opportunities that human analysts might miss.

4. Reduced Operational Burden: The laborious task of creating, maintaining, and debugging complex rule sets is largely eliminated, freeing up marketing and IT teams to focus on higher-value strategic initiatives.

The Crucial Role of Confidence Scores and Guardrails

One of the most powerful aspects of predictive AI is its ability to assign a confidence score to its predictions. The AI doesn't just say, "Send this." It says, "I'm 92% confident that Customer A will click on an email about Product C, but only 30% confident Customer B will open a promotional email today."

This confidence score is a game-changer because it allows us to set intelligent guardrails:

• Confidence Thresholds: We can decide to only send an AI-personalized email if the AI's confidence score for a positive engagement (e.g., open, click, conversion) is above a certain percentage (e.g., 70%). If the confidence is lower, we can default to a more general email or a human-curated message. This prevents "bad" or irrelevant recommendations from going out.

• Business Rules (Blacklists/Whitelists): While AI is powerful, human oversight is still critical. We can implement strict "blacklists" (e.g., never recommend a product a customer just returned) and "whitelists" (e.g., always send a welcome series to new subscribers). These are essential for brand safety, compliance, and ethical AI use.

A Salesforce Perspective: Streamlining the Tech Stack

For those of us entrenched in the Salesforce ecosystem (Marketing Cloud Engagement, Data Cloud, Marketing Cloud Personalization, Einstein Content Selection), this shift has profound implications, particularly for reducing AMPscript complexity.

Today, our email templates are often laden with dense, nested AMPscript for "if/then" logic. With AI, AMPscript transforms from a complex decisioning engine into a cleaner, more efficient tool for rendering dynamic content supplied by the AI.

• Less Logic, More Efficiency: Instead of hundreds of lines of AMPscript deciding what content to show, our scripts primarily focus on how to display the AI-determined content.

• Data Cloud as the Brain: Salesforce Data Cloud performs the heavy lifting of unifying customer data and deriving intelligent insights, which then feed into Marketing Cloud Engagement, vastly simplifying the AMPscript needed.

• Einstein Content Selection's Power: For dynamic content blocks, Einstein Content Selection eliminates the need for AMPscript to make content choices, as it makes these decisions in real-time based on AI predictions.

This results in cleaner, more readable AMPscript, faster email development cycles, and improved email performance. It's a move away from brittle, template-level logic to a centralized, intelligent AI layer.

The Path Forward: A Collaborative Effort

Transitioning to predictive AI isn't about replacing human expertise; it's about augmenting it. It requires close collaboration between marketing and IT. Our IT teams will be crucial in:

• Building robust data pipelines within Salesforce Data Cloud.

• Implementing and managing the AI/ML platforms.

• Defining and enforcing the critical guardrails and confidence thresholds.

• Monitoring AI performance and refining models.

By embracing predictive AI, we're not just improving our email personalization; we're investing in a scalable, adaptive, and truly customer-centric approach that will drive significantly higher engagement, satisfaction, and ultimately, business growth.

What are your thoughts on the shift to AI-driven personalization? Share your experiences in the comments below!

Predictive AI: The "Chicago" of Artificial Intelligence – Building the Future, One Prediction at a Time

 When you hear "Artificial Intelligence," what's the first thing that comes to mind? For many, it's the flashy, creative power of generative AI: think stunning AI art, captivating chatbot conversations, or even hyper-realistic deepfakes. These innovations are undeniably exciting, grabbing headlines and sparking imaginations. They are, in a way, the Hollywood of AI – dazzling, innovative, and highly visible.

But beneath the glitz and glamour, another, equally vital form of AI is quietly, relentlessly building the foundations of our modern world. It's the AI of "big shoulders," the tireless worker, the backbone that ensures everything runs smoothly and efficiently. This, my friends, is Predictive AI, and if generative AI is Hollywood, then predictive AI is most certainly the Chicago of Artificial Intelligence.

Why Chicago? Why "Big Shoulders"?

Think of Chicago's historical legacy, powerfully captured by Carl Sandburg in his iconic poem "Chicago":

• Industrial Might: "Hog Butcher for the World, / Tool Maker, Stacker of Wheat, / Player with Railroads and the Nation's Freight Handler;" Sandburg's lines evoke a city built on hard work, logistics, and the efficient movement of goods and people. It was the engine room, the processing hub, the city that got things done.

• Foundational Infrastructure: While other cities might boast of their cultural flair, Chicago was – and still is – a marvel of urban planning and infrastructure, connecting vast regions and enabling commerce.

• Reliability and Resilience: "Stormy, husky, brawling, / City of the Big Shoulders:" Through booms and busts, Chicago has consistently proven its ability to adapt, endure, and continue its vital work.

Predictive AI shares these same fundamental characteristics. It's not always the star of the show, but its power is immense, its impact far-reaching, and its role absolutely indispensable.

What Does Predictive AI Do? The Workhorse in Action

While generative AI creates new things, predictive AI excels at anticipating what will happen next based on vast amounts of historical data. It analyzes patterns, identifies trends, and forecasts outcomes, empowering businesses and organizations to make smarter, more proactive decisions. It is truly the "City of the Big Shoulders" for the digital age.

Here's where predictive AI flexes its "big shoulders":

• Optimizing Supply Chains: Imagine a world where every delivery is delayed, every shelf is empty. Just as Chicago was the "Nation's Freight Handler," ensuring goods flowed smoothly, predictive AI takes on the monumental task of optimizing supply chains. It analyzes weather patterns, traffic, demand fluctuations, and supplier performance to predict disruptions and optimize logistics, ensuring products get where they need to be, when they need to be there.

• Forecasting Demand: Retailers use predictive AI to anticipate exactly how many units of a given product will sell, preventing both costly overstocking and frustrating stockouts. This helps businesses manage resources, reduce waste, and maximize profits – pure efficiency, echoing the work of a "Stacker of Wheat."

• Preventing Equipment Failure: In manufacturing plants or critical infrastructure, unexpected equipment breakdowns can be catastrophic. Predictive AI monitors sensor data from machines to predict when a component is likely to fail, enabling proactive maintenance and preventing costly downtime. It's the ultimate "Tool Maker," ensuring the machinery of industry hums along.

• Detecting Fraud: Banks and financial institutions rely heavily on predictive AI to flag suspicious transactions in real-time. By identifying patterns of fraudulent activity, it acts as a silent guardian, protecting billions of dollars and countless individuals from financial crime.

• Personalizing Experiences: Ever wonder how Netflix knows exactly what movie you might like next, or how Amazon recommends products you actually want? That's predictive AI analyzing your past behavior and preferences to anticipate your future interests.

• Improving Healthcare: From predicting disease outbreaks to identifying patients at high risk of readmission, predictive AI helps healthcare providers make more informed decisions, leading to better patient outcomes and more efficient resource allocation.

The Unsung Hero: Why "Not Sexy" is Actually Profoundly Powerful

Perhaps predictive AI isn't "sexy" in the way a viral AI art generator is. You don't necessarily interact with it directly in your daily life, and its successes are often defined by things not happening – no delayed packages, no fraudulent charges, no unexpected machine failures.

But this quiet, consistent impact is precisely what makes it so profoundly powerful. Predictive AI doesn't seek the spotlight; it enables the systems that make our lives easier, our economies stronger, and our industries more resilient. It's the steady, reliable engine room of the AI revolution, making sure the entire ship stays afloat and moves forward. It works "under the terrible burden of destiny laughing as a young man laughs and eating up the years like a hungry fighter," as Sandburg put it, relentlessly building a more efficient and predictable future.

So, the next time you marvel at a piece of generative AI art, take a moment to appreciate its steadfast cousin. Predictive AI, the "Chicago" of AI, is busy building, optimizing, and predicting the future, one data point at a time – and that, in its own enduring way, is incredibly impressive.

Underground Leaders: Why Gen X Is Quietly Powering the AI Marketing Revolution

It was a classic Chicago spring afternoon—clear skies, a good game, and then suddenly… a sharp wind off the lake. We were in a suite at Wrigley Field for a corporate marketing event, enjoying the game from the balcony seats, when the temperature dropped nearly 20 degrees in what felt like five minutes. That’s early May in Chicago for you. As everyone filed back inside for warmth, the shop talk began.

A few of us—Gen Xers who are now leaders on major corporate marketing teams—started comparing notes. Not on the weather, but on how we manage. How we lead differently than the generations before us. The conversation quickly shifted to transformation, risk, and how we’re approaching AI—not as a buzzword, but as a toolset.

That conversation stuck with me.

It reminded me that, just like we did in the early days of our careers, Gen X is once again working just outside the spotlight. But make no mistake: we’re driving the real transformation. Especially when it comes to AI in marketing.

---

From the Margins to the Middle Seat

Back in the '80s and early '90s, I was grinding my way through undergrad at Saint Louis University while working full-time. It took me eight years to finish, but I didn’t see that as a failure—I saw it as survival. My friends and I weren’t tuned into the mainstream. Our culture lived off the dial.

One of my closest friends used to order import records from the UK and loan them to DJs at WMRY 101.1 FM, a progressive rock station out of Belleville, Illinois. WMRY was unlike anything else in the market—owned by the Catholic Shrine of Our Lady of the Snows, but with no real religious programming. It was commercial-free, playing alternative and underground music, and the DJs programmed their own shows. It was pure Gen X: DIY, under the radar, and full of signals the mainstream hadn’t caught yet.

In contrast, KSHE ruled the St. Louis radio market with a Boomer-focused classic rock format. We didn’t fit there. And that’s the point.

That yellow WMRY bumper sticker? It’s still taped to the wall of my old room at my parents’ house. I see it every time I visit. It reminds me that we weren’t raised to follow trends—we were raised to find what mattered and amplify it ourselves. That mindset never left. It just evolved.

We grew up finding our own path—whether it was music, media, or work. That mindset didn’t go away when we got promoted. It’s what’s powering how we lead today.

---

Why Gen X Is Built for AI Leadership

AI adoption in marketing doesn’t succeed because you added “machine learning” to a slide. It succeeds because someone—usually behind the scenes—is testing, iterating, refining, and proving real value. That’s where Gen X shines.

We're the ones who:

• Evaluate tools not by hype, but by lift.
• Ask the awkward questions early.
• Launch test pilots before asking for budget.
• Measure performance, not impressions.

We came of age learning to fix our own problems. We don’t expect AI to be magic. We expect it to be useful.

---

AI Isn’t Aspirational—It’s Operational

In corporate marketing departments, there’s often a lot of energy spent on AI optics. Executive decks, vision statements, innovation summits. But while that theater plays out, Gen X leaders are in the trenches:

• Setting up predictive models in Salesforce and actually checking the regression strength.
• Using ChatGPT to automate creative briefs—not to write the campaign.
• Testing dynamic content frameworks that reduce review cycles by 40%—not just adding personalization tokens and calling it a day.

We’re building systems. Not hype.

---

Bridging the Generational Divide

We also play a unique role in today’s multigenerational workforce. We understand the Boomer emphasis on structure and reliability. We appreciate the Millennial and Gen Z desire for transparency and flexibility. And because we’ve had to operate in both worlds, we know how to translate between them.

That makes us natural integrators—a vital role as AI touches every part of the marketing org, from creative to ops to compliance.

---

The Bottom Line

Gen X leaders aren’t waiting to be anointed the “AI transformation generation.” That’s not how we work.

We’re shipping MVPs. We’re building content ops. We’re automating backend workflows so creative teams can breathe again. We’re not reinventing marketing. We’re making it work better—with AI as a tool, not a trophy.

We didn’t grow up expecting a seat at the head of the table. But don’t mistake that for passivity. We’re not the loudest voice in the room—but when the AI revolution in marketing hits its stride, look around.

Chances are, a Gen Xer built the engine.

---

Let's Talk:

Are you a Gen X marketing leader quietly rolling out AI inside your org? Or someone who came up through the same analog-to-digital shift with an eye for impact over applause? I’d love to hear your story—and how you're putting AI to work without the theater.

The Kid’s Table Isn’t a Punishment—It’s a Prototype Lab

While the grown-ups were busy chasing shiny GenAI demos and making splashy announcements at board meetings, something interesting was happening in the corners of the enterprise. The teams without headcount. The projects without press releases. The initiatives not on the CEO’s radar.

Welcome to the kid’s table.

For anyone who’s worked in a large company, you know the feeling. You’re not at the “main table” where the enterprise-wide AI vision is being defined. You’re not the face of the transformation initiative. You're just the person in the back, trying to do something useful with the tools already in front of you—like a dusty personalization model or an underused CDP license.

But here’s the thing: the kid’s table is often where the real innovation happens.

Because nobody’s watching too closely. And that’s your advantage.

Why Innovation Thrives at the Margins

When you’re outside the spotlight:

• You can experiment without being micromanaged.
• You’re not locked into executive-imposed deadlines or artificial KPIs.
• You can build for real impact—not just optics.

This is how real transformation often begins inside big companies. It doesn’t start with a 100-slide PowerPoint. It starts when someone quietly builds a working prototype using tools the company already owns—and proves that it drives results.

These “skunkworks” projects don’t get budget—at least not at first. What they get is permission through obscurity. And that’s often all you need.

From Skunkworks to Superstars: Real Impact from the Edge

History is filled with examples of big companies giving small, autonomous teams the space to operate like startups—and reaping huge rewards in the process.

• Lockheed Martin’s Skunk Works: Developed the U-2 and SR-71 in secret, pioneering stealth technology and setting a gold standard for agile innovation under pressure.
• Apple’s Macintosh Team: A renegade group launched one of the most disruptive personal computing platforms ever—paired with the “1984” Super Bowl ad that redefined tech marketing.
• IBM PC Division: Escaping corporate inertia, this team created the IBM PC in record time, establishing the architecture that dominated computing for decades.
• Toyota’s Prius Project: Despite internal skepticism, a small team launched the first mass-produced hybrid vehicle, creating a category and boosting Toyota’s green credentials.
• 3M’s Post-it Notes: A side project from an adhesive experiment turned into an iconic product—thanks to internal evangelism and bottom-up marketing.
• Google X (now X, a Moonshot Factory): Created industry-shaping innovations like Waymo and Project Loon—projects too risky for the main org to touch.

In many of these cases, what started as “off to the side” experiments—hidden from corporate micromanagement—eventually transformed not just product portfolios, but company cultures and market leadership positions.

Agile Teams in Legacy Giants: Not Just for Tech

It’s not just the Apples and Googles of the world. Even deeply established, non-tech brands have built agile, startup-style teams to break free from red tape:

• John Deere’s XI teams slashed innovation cycles from 3 years to 8 months—and saw a measurable rise in team morale and delivery quality.
• Roche Korea empowered patient-focused agile squads that delivered a 30% boost in sales and improved outcomes in a year.
• Constellation Brands’ Mission Bell Winery used agile pilots to 10x problem-solving speed in its distribution unit.

These efforts weren’t announced on stage with fanfare. They started by trusting small teams to operate with speed, creativity, and autonomy—and they worked.

But Let's Be Honest: Not All Skunkworks Succeed

A word of caution: the “kid’s table” can turn into an unused playroom if the organization only supports innovation in theory. Skeptics rightly point out common failure modes:

• Innovation Theater: Projects launched for optics, not outcomes
• Integration Wall: Breakthroughs that can’t scale or fit into the mothership
• Cultural Backlash: Perceived favoritism or “special teams” that erode trust

That’s why these efforts must be grounded in customer problems, tied to clear outcomes, and championed by leaders who protect the space to build—even when the results aren’t immediate.

Acting Like a Startup Inside a Corporation

If you’re trying to shed the perception that you’re “too corporate” for startup life, here’s the truth: The best startup operators in big companies already act like founders. They:

• Navigate constraints instead of complaining about them
• Use existing resources in new ways
• Build lean MVPs before asking for permission
• Understand how to scale only after product-market fit is validated

Working from the margins is not a weakness. It’s training.

Closing Thought: Reclaiming the Kid’s Table

As we wrap this “AI at the Kid’s Table” series, I’ll leave you with this:

Being at the kid’s table doesn’t mean you’re immature or irrelevant. It means you’re not stuck defending the status quo. You have permission to play, to prototype, and to break the rules just enough to build something better.

And sometimes, the biggest AI breakthroughs don’t come from the teams with the biggest budgets. They come from the people who stopped waiting for permission—and just started building.

AI Isn’t Magic. It’s Math. Why Marketing Execs Miss the Real Power of Tools Like Einstein

If you’re leading a marketing team in 2025, chances are you’ve been pitched AI from every direction: image generators, chatbots, virtual influencers. You’ve probably seen demos of dazzling creative tools that promise to slash costs or rewrite campaigns in seconds. And maybe—just maybe—you’ve assumed that your Salesforce investment already has “AI” baked in, and it’s doing its thing behind the scenes.

But here’s the problem: Most organizations aren’t actually using the AI they already have. And worse, many marketing leaders still misunderstand what AI is—and what it isn’t.

1. AI Isn’t Magic. It’s Modeling.

Too often, AI is perceived as a black box—a magical engine that just knows what to do. But in reality, AI is far more mundane—and far more powerful when treated accordingly.

At its core, AI in marketing is about probability. Predictive models use historical and real-time data to make statistically informed guesses:

• Who’s likely to open this email?
• Which image will get the most clicks?
• What time should we send this message?
• Which product should we recommend next?

It’s math, not magic. And treating it like a wizard instead of a workhorse is one reason so many “AI transformations” stall out.

2. The Problem with Chasing Hype: Generative Gets the Buzz, Predictive Gets Results

Right now, generative AI gets all the attention. From executives wowed by AI-generated photos to internal teams experimenting with ChatGPT for copywriting, it feels like the future is here.

But most of the business value in AI today lies in predictive tools—not creative ones.

Global adoption of generative AI technologies—including text, image, and chatbot generation—has risen dramatically, increasing from 33% in 2023 to 71% in 2024 according to McKinsey. However, only around 10% of companies have fully integrated these tools into core operations (McKinsey, 2024).

By contrast, personalization and predictive AI tools—such as Salesforce Einstein—have seen slower adoption. Reports indicate only 39–59% of Salesforce Marketing Cloud customers actively use features like predictive scoring, content selection, or send-time optimization (Damco, 2025 | Redress, 2025).

Put simply: Generative gets headlines. Predictive drives revenue.

3. What Salesforce Einstein Actually Does (That Your Team Might Be Ignoring)

Salesforce’s Einstein brand includes a broad set of AI features, most of which support automation, personalization, and optimization—not image generation or chatbot content.

• Einstein Content Selection (ECS): Swaps in personalized banners, CTAs, or promotions based on individual-level data.
• Einstein Engagement Scoring (EES): Prioritizes active, high-intent users in your CRM.
• Send Time Optimization (STO): Dynamically staggers send times for each person.
• Predictive Lead Scoring: Improves sales and dealer handoff strategies.

Yet Salesforce’s own ecosystem data shows that fewer than 40% of Einstein-enabled organizations use more than one of these features—and that usage is rarely integrated across customer journeys.

4. Creative Bias: The Quiet Killer of AI ROI

In many brand-driven organizations, creative still holds disproportionate influence over campaign direction.

That’s not a critique of design. But when aesthetics take precedence over testing, targeting, and decisioning, the full value of AI is often left untapped.

Classic direct marketing wisdom: 60% of success comes from the list, 30% from the offer, and only 10% from the creative.

Contemporary data backs this up:

• Adobe’s 2024 Digital Trends report found that dynamic personalization strategies drove 2–3x more revenue per send than campaigns emphasizing creative refreshes alone.
• A Fortune 100 retailer discovered that triggered emails using Einstein Content Selection generated 4x the revenue of their highest-profile seasonal campaigns—but struggled to secure sustained executive support.

5. You’re Already Paying for AI—Even If You Don’t Know It

Whether you're at a global brand or a scrappy startup, chances are you’re already paying for AI.

In enterprise platforms, tools like Salesforce, Adobe, and HubSpot bundle predictive AI features into their licensing fees. But platform inclusion does not equal operational impact.

Instead, AI is frequently treated as a compliance checkbox:

• ✔️ AI-enabled CRM
• ✔️ AI-ready platform
• ❌ Actually driving business impact

Salesforce CEO Marc Benioff recently positioned Agentforce—its next-gen AI framework—as the centerpiece of the company’s future, noting on CNBC:

“We have pivoted our company hard and fast... The power of Agentforce is that our apps, agents, and data cloud are all on one unified platform. The agents are fluid, available anytime, in any app, in the flow of work.”

It’s an ambitious vision—but one that risks being undermined by foundational gaps in adoption. Many organizations haven’t even fully leveraged baseline features like STO or EES, much less deployed agentic AI workflows.

6. The Fix: Start with Strategy, Not Shiny Objects

If you’re a marketing leader, the real unlock isn’t another AI pilot. It’s integration.

• Start with a business question: How can we recover more cart abandons? How can we upsell service plans in the ownership phase?
• Work cross-functionally to connect CRM, content, and analytics.
• Shift creative thinking toward modular, testable assets.
• Measure incremental lift from personalization, not just campaign performance in aggregate.

AI isn’t a strategy—it’s an amplifier. If your core execution is sloppy or misaligned, AI will magnify that dysfunction. If your processes are disciplined, AI can generate real operational lift.

Closing Thought

AI in marketing isn’t about creating more content. It’s about creating better outcomes.

If your team is obsessed with campaign visuals but ignoring send time optimization, engagement scoring, or dynamic decisioning, you’re not leveraging AI—you’re decorating it.

Now is the time to move beyond the hype and start realizing the potential of the tools already within your reach.