Blog
Three customer loyalty and retention strategies that build long-term loyalty
Learn more
Loading Insights...
Artificial intelligence is everywhere in marketing right now.
Ninety-three percent of marketers say they’re allocating at least 5% of their budgets to AI initiatives. Nearly half consider AI capabilities an extremely important factor in vendor selection. Predictive models, decisioning engines and personalization algorithms are being positioned as the answer to everything from media efficiency to customer loyalty.
The excitement is justified. AI can help brands decide who to reach, what to say, where to show up and when to engage. But there’s a growing disconnect between AI ambition and AI performance. Because in advertising, AI doesn’t fail due to weak models. It fails due to weak identity.
This isn’t about generative tools that produce copy or images or about surface-level automation. It’s about the AI that decides, optimizes and measures marketing performance over time, which depends on understanding behavior across interactions. And that foundation lives in the data and identity layers.
As Digiday recently reported, most platforms buy against broad contextual signals and layer targeting on top, while Epsilon has taken a different approach—building its AI around identity from the ground up.
Predictive AI systems are designed to learn from patterns over time. They forecast outcomes based on historical behaviors and contextual signals. But learning requires stable intelligence.
Without persistent identity resolution, a customer quickly becomes a collection of disconnected signals:
Each interaction exists, but the person behind them does not.
When that happens, AI defaults to optimizing fragments instead of customers. Models over-weight recent clicks. Attribution skews toward easily observable channels. Lifetime value becomes an average instead of a trajectory.
The issue isn’t that the algorithms aren’t sophisticated enough. It’s that the data beneath them isn’t unified. This is where a customer data platform (CDP) becomes foundational.
An enterprise CDP cleans, unifies, completes and enhances customer data to create a persistent, cross-channel profile. When a CDP has identity resolution built in, it aligns online and offline signals to real individuals and households. It can ensure duplicate records are consolidated and contact information gaps are filled.
In addition, external demographic and intent data can be added to enrich unified profiles for greater understanding of individuals. So instead of teaching AI from impressions and events, brands can teach it from people. And that changes everything.
In advertising, AI ultimately powers four core decisions:
Those decisions span paid and owned channels. They require coordination across display, social, email, SMS, direct mail, website personalization and more.
Without identity resolution, those channels operate independently. Paid media knows what was clicked. Owned channels know what was purchased. Social knows what was shared. But no system understands how one person engaged across all three.
The result is over-frequency and disjointed touchpoints. The fragmented experience this produces is what drives inefficient spend.
A CDP anchored in person-based identity creates a single, persistent customer ID that travels across activation channels. With solutions like Epsilon CDP, marketers can unify CRM records, website behaviors, transaction history and media exposure into one profile.
That profile then becomes the foundation for activation across owned and paid channels.
The shift is subtle but important. Instead of activating audiences defined by one channel, brands activate real people recognized across channels. AI moves from optimizing impressions to orchestrating customer journeys.
AI-powered advertising increasingly operates in real time with offers changing dynamically, website content adapting in-session and media pacing adjusting based on performance signals.
But real-time decisions require real-time recognition.
If a system cannot recognize a returning visitor or connect a recent transaction to a current browsing session, the decision engine is working in the dark.
Epsilon’s COREai is designed to make individual-level marketing decisions, determining the best target, best offer, best channel and best timing. But it does so within the context of a unified identity layer. Real-time recognition connects incoming signals to an existing profile. Automated model refreshes ensure predictions reflect the most current behaviors.
Without that persistent profile, “real-time AI” becomes reactive guesswork. With it, decisioning becomes contextual and measurable.
If AI is becoming the control system of marketing, identity is the operating system beneath it.
The intelligence layer cannot function without a stable data layer.
AI doesn’t just need to decide. It needs to improve. And improvement requires closed-loop measurement.
Traditional measurement approaches often separate media exposure from transaction data. Impressions may be anonymous. Site visits may not resolve to known individuals. Conversions may live in a different system altogether.
When identity is fragmented, attribution models rely on proxies. Channels appear more or less effective based on what can be observed, not what actually influenced the customer.
A CDP grounded in identity brings together interactions based in personally identifiable information and pseudonymized media signals into a unified profile. It enables brands to link exposure, engagement and transaction at the person level—not just the campaign level.
With person-based, closed-loop measurement:
AI that cannot measure at the customer level cannot optimize at the customer level. Identity is what makes optimization durable instead of directional.
As AI investment accelerates, the conversation is shifting from tools to infrastructure. Enterprises aren’t struggling with too little data. They’re struggling to act on it before the moment passes. They’re navigating expanding privacy regulations, walled gardens, and growing channel complexity.
An AI-ready data layer requires:
Not every CDP includes robust identity resolution. And not every identity solution is built to operate at enterprise scale.
Epsilon CDP was built with person-based identity at its core, enabling privacy-forward activation and measurement across owned and paid channels. With deterministic matching and persistent IDs, it supports the kind of AI-driven decisioning modern advertising demands.
Brands that win in AI-driven marketing won’t simply have the best models. They’ll have the most trusted and persistent identity foundation.
To see how this plays out in practice, consider how AI operates in automotive brand marketing with and without a persistent identity layer. When optimization is built on fragmented or limited first-party data, models are refreshed manually—sometimes every six weeks—and operate on broad audience snapshots like “auto intenders.” Decisions are reactive and generalized.
By contrast, when AI is powered by a CDP anchored in person-based identity, models refresh in real time and recognize in-market individuals as they engage. Instead of targeting a generic segment, the system can identify a specific SUV buyer based on recent behaviors and purchase history, enabling the adjustment of messaging, channel and timing accordingly.
That shift from audience-level to individual-level decisioning acts as a force multiplier because auto brands can identify and engage two to three times more in-market customers because they’re no longer guessing at segments but recognizing real people across channels.
Think of AI’s function here as a layered system. Specialized machine learning models handle real-time, high-frequency decisioning, while LLMs can support more language-driven tasks like reporting and workflow operation. As Digiday noted in its reporting on Epsilon’s approach, this kind of system relies on multiple models working together, with decisioning models at the core, LLMs supporting them and human oversight where precision matters most.
AI is reshaping advertising. Predictive models are improving targeting. Real-time decisioning is accelerating personalization. Investment in AI continues to rise. But intelligence doesn’t exist in isolation. It depends on memory, context and continuity. It depends on identity.
For AI to work in advertising, the underlying data architecture has to be built on identity. If the control system of marketing is becoming AI, then the operating system beneath it must be a unified, identity-driven data foundation. And that foundation lives in the CDP.
If you’re evaluating how AI fits into your advertising strategy, start by examining the layer beneath it. Because the quality of your identity resolution may determine the ceiling of your AI performance. Learn more about the power of Epsilon CDP.
