The agentic marketing stack starts with the data layer

The agentic marketing stack starts with the data layer | Databricks Blog Skip to main content

Summary

Acxiom is building an end-to-end agentic marketing value chain — from identity resolution and audience planning to media buying, campaign activation and performance analytics — and the foundation making it possible is a modern, cloud-native data architecture.

Workflows that used to require months of manual effort from data engineers, architects and creative agencies are now being prototyped in hours; Acxiom has seen performance improvements of 80 to 90 percent when migrating from on-premises data center to Databricks.

The shift from delivering data through file transfers to embedding it agentically inside client environments is redefining Acxiom's competitive position — from data supplier to intelligence layer inside the marketing stack.

There's a version of the AI modernization story that goes: build the platform, then figure out the use cases. Ankur Jain would tell you that's backwards — and that most organizations are learning that the hard way.

Ankur is Chief Cloud and Data Modernization Officer at Acxiom, the connected data and technology foundation that helps global brands resolve customer identity across channels, enrich customer profiles with more than 10,000 attributes, and deliver outcomes across customer acquisition, retention and personalization.

Ankur leads both product engineering and client-facing solutions engineering — meaning he is responsible not just for what Acxiom builds, but for how those capabilities get embedded inside the environments where clients actually operate.

After joining the company less than two years ago, Ankur led the modernization of Acxiom’s core infrastructure, data pipelines, legacy architecture and underlying tech-stack. Today, Acxiom is actively building agentic workflows that automate the full marketing value chain.

Why the Foundation Has to Come First

Aly McGue: A lot of organizations want to move to agentic AI but are still running core workloads on legacy infrastructure. What is the risk of trying to build intelligence on top of a foundation that wasn't designed for it?

Ankur Jain: The risk is that you hit a ceiling almost immediately. When I joined Acxiom, both products and client solutions were hosted mostly on-premises. When your products and solutions are constrained to a data center, they have limited scalability. Performance was not up to par for the real-time use cases clients were asking for. And then there was a lot of legacy tech — the stack needed a refresh, a reimagining of what cloud-native architecture could look like.

What we also saw was a lot of manual pipelines, a lot of data redundancy, copies of the same data in multiple places. The process itself was not very efficient. Any organization trying to build agentic capabilities on a fragmented or legacy foundation is going to spend more time managing infrastructure than building products.

For us, the strategic vision comes down to two north stars: data modernization and agentic marketing. They are sequential, not parallel. You cannot build an agentic marketing ecosystem on a legacy foundation.

How a data warehouse migration shifted the focus from maintenance to business outcomes

Aly: You moved from on-premises Hadoop to Databricks. What did that shift make possible that wasn't possible before?

Ankur: In terms of performance, we have seen improvement across the board, across different types of workloads and different types of pipelines, almost 80 to 90 percent faster run times. Workloads that used to take 50+ hours, sometimes 90+ hours — and I'm talking hours, so literally days, sometimes up to a week — are now getting done within 2-3 hours. Those same workloads, in 2-3 hours.

It has also freed up our people. In some cases we have been able to free up multiple full-time roles to focus more on value-added outcomes rather than managing infrastructure. The number one thing it enabled was for the engineering team to focus more on business outcomes rather than worrying about the infrastructure underneath. That might sound like a soft win, but when your engineers are spending their time building products and delivering client solutions rather than keeping the lights on, it changes what you can even attempt.

What the Agentic Marketing Value Chain Actually Looks Like

Aly: Where are you seeing agentic AI reshape actual marketing workflows today, and where does that vision extend?

Ankur: Acxiom's core operation is very data-centric. We bring in marketing data from multiple platforms — CRM, e-commerce, Adobe Analytics, Google Analytics — and help brands build a holistic customer view, enrich it, and deliver outcomes. Traditionally, that required a team of data engineers and data architects who would model everything and build pipelines manually. ETL is always the longest pole in the tent, and it would take months.

Through AI, that entire cycle compresses. Code generation through prompts, automated testing of outputs, accelerated CI/CD pipelines. On the marketing side, producing different variations of an ad used to take creative agencies months. Now you can analyze ads at scale through machine learning, feed those results into an AI engine and generate highly customized variations in minutes.

Where we have seen the biggest real shift is on execution. Take audience planning — a marketer passes a prompt describing a campaign objective and target profile, and the agent builds the audience segments with sample personas using Acxiom data, surfaces different demographic and behavioral dimensions and lets the marketer refine from there. What used to take effort from multiple people with varied skill sets and a lot of lead time is now done agentically in minutes. We have demonstrated the same pattern for media buying: an agent queries available inventory, evaluates it, makes a buying decision and activates the audiences across channels.

The goal is to connect the entire pipeline — from audience design through media buying, activation and performance analytics — into an agentic framework. That whole AI for BI capability that Databricks is building through the Genie and agentic ecosystem is exactly where marketing workloads like ours are heading. It can all be put to work end-to-end.

How governance accelerates agentic workflows

Aly: Acxiom operates in highly regulated industries, and deploying agents requires a high level of...