An Agentic CDP is built for the AI era. It powers always-on 1:1 personalization with Infinity Campaigns, is embedded in the data foundation where governed customer context lives, and is architected from the ground up for agents as first-class operators
by Tasso Argyros, Ali Ghodsi and Reynold Xin Marketing technology has seen plenty of change over the past few decades. But what is happening right now is different, because two massive shifts are arriving at the same time.
The technology stack marketers rely on is being disrupted by AI and agents, and the behavior of the modern buyer is changing, because of AI and agents, in ways that make the old stack incompatible with what comes next.
The traditional Customer Data Platform (CDP) is not just becoming dated, but is entirely the wrong tool for today’s job.
Picture a customer who wants to book a flight. Within seconds, she deploys three agents: one to research routes and airlines, one to scan her inbox for loyalty offers and compare prices across the web, and a third to make the purchase and negotiate the best deal. What used to be a multi-week journey now happens in milliseconds, across every channel at once, without a human consciously steering it.
This creates three requirements that today's marketing infrastructure was never built to meet.
The first is speed. Agentic buying lifecycles operate in milliseconds. Any system working in batch cycles of days or weeks is invisible to this buyer by the time it responds.
The second is hyper-personalization. Agents are efficient filters. Marketing content that isn't precisely relevant to this customer, at this moment, gets discarded before any human sees it. A first name in a subject line or a segment of ten thousand people isn't personalization by this standard.
The third is richer context. Marketing teams have long relied on data engineering to assemble a "Customer 360" dataset for campaigns. Agents will generate more signals than ever, but raw data isn't the bottleneck anymore. Agents need context: a live picture of the customer, the business, and the history of every decision made about that customer and why.
The industry has spent years building toward the "Golden Record" — what many teams call Customer 360. It's a unified customer profile: demographics, transactions, behavioral history. It answers the question of who the customer is, and it's genuinely useful.
Agents need more than that. They need Golden Context, which adds two things the Golden Record doesn't carry: what the business is trying to accomplish right now, and what's already been tried with this customer and how they responded.
The airline example makes this concrete. An agent working from a Golden Record knows a customer flies eight times a year and holds elite status. An agent working from Golden Context knows that the same customer's flight is delayed by two hours, that she's traveling with three kids under ten, that the lounge has capacity for four tonight, and that six months ago she complained about a delay and nobody followed up. The first agent sends a generic loyalty email. The second one fixes the trip.
The shift to agents and context affects every tool marketing has been using, but maybe none as much as the CDP.
Historically, CDPs have served as the vital middleware of the marketing stack, sitting between data platforms and execution tools to allow human marketers to organize and distribute data for batch-based audience targeting. However, the emergence of "agentic buying", where autonomous agents research and transact on behalf of humans, renders this middleware architecture obsolete.
Because traditional CDPs were built for static, rule-based campaigns managed by humans, they fail to meet the three critical requirements of the agentic era:
Regardless of the CDP flavor of your choice, “bundled” or “composable": these systems were simply not designed for an era of agents researching and buying on behalf of humans, nor are they native to modern agent-first architectures.
A new generation of CDP needs to serve the needs of Marketers for the next decade and beyond. One that uses agents at the heart of its architecture; gets rid of the concept of large audiences and static rules in favor of intelligent, agentic responses to customer signals as they happen; and helps not only with the campaigns but also helps collect and organize the data & context needed to drive that intelligence.
An Agentic CDP is built for the AI era. It powers always-on 1:1 personalization with Infinity Campaigns, is embedded in the data foundation where governed customer context lives, and is architected from the ground up for agents as first-class operators alongside humans.
The next generation of CDP needs to be built with agents at the center of the architecture, not bolted-on as a feature after the fact. An Agentic CDP has three defining characteristics:
We need a new engagement concept for this era – beyond the campaigns and journeys of old. If engagement must be continuous and not episodic, if it needs to react with the speed of agents that think in milliseconds, and personalize to segments of one – a single human and her agents, then this is a materially different thing than a campaign.
Legacy CDPs were built to fuel batch and blast marketing campaigns that are slow, static, and could never truly be 1:1. Even customer journeys are highly manual and rule-based, lacking true personalization at the individual level.
The Agentic CDP, for the first time, offers a completely new capability we call Infinity Campaigns: autonomous, continuously compounding engagement loops that constantly adapt to new context signals to autonomously reshape their message, timing, and channel on the fly.
In other words, Infinity Campaigns are:
To revisit the airline example – A marketer can easily build a campaign for a loyal traveler who flies more than eight times a year. The marketer will use demographics, transactions, campaign history, and engagement records to “personalize” a timely message inviting the traveler to an exclusive offer for their next booking.
But what if that loyal customer’s flight just got delayed? They’re traveling solo with three kids under ten and frantically asking the airline app’s chatbot if their airport has a restroom with a changing room. A generic booking email isn’t the right move.
An Infinity Campaign should be always-on for that customer. By processing and understanding all of those signals, the Campaign Agent can determine that the best action is to host the family with complimentary lounge access and rescue their brand experience.
The Agentic CDP is designed from the ground up to provide this core capability, natively.
The Agentic CDP lives inside the data platform, not alongside it.
This follows from what Golden Context actually requires. Customer context, business context, and decision context all live in the enterprise data platform. Agents need fast, secure access to all three at once. A CDP sitting outside the data platform, pulling data across an integration, will always be too slow and too incomplete for agents running in milliseconds.
The practical result is a collapse of the layers of the traditional martech stack. The lakehouse and the CDP need to occupy the same layer so that context is collected, managed, and analyzed in one place rather than synchronized across systems with the inevitable speed, security and governance challenges that would stem from that separation.
Governance matters here too, and not as an afterthought. CDPs touch some of the most sensitive data enterprises hold: PII, behavioral records, transaction history. The governance systems protecting that data (and increasingly agents) live inside the Lakehouse, using innovations like Unity Catalog. An Agentic CDP integrated with that governance layer means every agent action runs under the same data entitlements and security boundaries as every other operation in the enterprise. No separate ruleset for the CDP, no delays standing up new use cases, no parallel systems to maintain.
Finally, Golden Context is only as good as the data behind it. When the CDP is Embedded in the Lakehouse, truly difficult problems such as resolving customer identities across channels or catching and correcting data quality issues, are solved in the same place the data already lives.
Every CDP that existed before the dawn of ChatGPT was designed for human operators: dashboards, logins, manual rules, scheduled jobs. Some vendors have since layered in agentic features, and some of those features are genuinely useful. But there's a meaningful difference between adding a chat interface to an existing system and building a system where agentic operation is the primary mode from the start. The former is an upgrade; the latter requires a rebuild.
The Agentic CDP built for the LLM era is designed so that every function can be run by an agent or a human. Humans still set goals, review outcomes, and stay in control, but the system moves beyond a world where humans have to manually build journeys step-by-step.
The marketing infrastructure that worked for the last decade is now mismatched with the buyers, channels, and speeds of the next one. The brands that win the next decade won't be the ones who got the best out of the old model. They'll be the ones who can meet customers (and their agents) where they actually are, at the moment it matters, at a scale that wasn't previously possible.
That's why we at Databricks built CustomerLake: a native Agentic CDP for the Databricks platform.
CustomerLake is built to deliver on the three principles above:
1: Infinity Campaigns as the core engagement model
2: Embedded natively in the Databricks platform
3: An architecture where agents and humans work together from day one, rather than agents being bolted on later
If your team is ready to stop building campaigns for a buyer that no longer exists, CustomerLake is built for what comes next.
To learn more, read our [product announcement blog](https://www.databricks.com/blog/introducing-customerlake-agentic-cdp).
Subscribe to our blog and get the latest posts delivered to your inbox.