As VP of data and analytics at Atlanta-based Georgia-Pacific, Matt Robuck gives a multitude of insights to the business. In the past, these would’ve been delivered as dashboards and reports, but over the last three years, with the proliferation of AI, he and his team of over 180 engineers have been questioning how they can strategically implement emerging tech to solve problems differently.
“Georgia-Pacific has been around for nearly 100 years, and many of the problems we’re trying to solve today we’ve been dealing with for decades,” he says. “New technologies allow us to look at these challenges through a different lens.”
Metadata and the AI ecosystem
According to Robuck, developing an AI ecosystem starts with having a very strong bedrock of data. “We’ve made substantial investments over the past three years to build these foundations,” he says. “In the past, data quality, metadata and contextual data weren’t seen as critical because if something was wrong in a report, a human would usually catch it. But with AI agents, there usually isn’t a human in the loop until much later in the process, which means poor-quality data can have big consequences.”
When building the data platform, Georgia-Pacific invested in data lake architecture, proactive data quality, and metadata. “We couldn’t have moved as fast as we have without these things,” he adds. For Robuck, the metadata component — the data context — is critical to the success of their AI agents.
With over 1,000 applications running across the company, each one has its own naming conventions, data types, and data structures. So when an agent is put on top of these data sets, it’ll do its best to interpret the data. “But we need more deterministic outcomes,” he says. “Our agents need to be right more often than not. Metadata, therefore, acts as a translation layer, giving the agents critical context.”
And by making the metadata even richer, with more detailed information about Georgia-Pacific’s different manufacturing sites, like size, products they produce, and business lines they serve, the agents have deeper knowledge of what the data means for the business more broadly.
The company considered building their own metadata solution, but given its size and scale, it partnered with agentic data intelligence platform Alation to ensure the process was quick and cost-effective.
Using data to solve real business problems
One area where these foundations are now being used to add value is across the business’ marketing function. Georgia-Pacific invests heavily in both digital and traditional marketing to promote its consumer products. But when a business wants to understand if its marketing budget is being allocated effectively, it must be able to track campaign performance.
“If you budget $50 million for a campaign, but no one sees it, that’s a nice waste of money,” Robuck says. The marketing team used to spend months at a time consolidating data to determine whether a particular campaign performed well, and then adjust strategies based on this deep dive.
“Understanding marketing performance is ultimately a data challenge,” he says. “That’s why we built a solution that pulls near real-time data from multiple marketing platforms, and turns it into actionable insights, giving marketing teams a far clearer view of how their campaigns perform.”
This strategy cuts costs, eliminates about 30,000 man hours every year, and helps the business connect with a wider customer base. In addition, if the marketing team deploys a campaign that isn’t performing well, having near real-time data allows them to pivot before they’ve wasted time and money on the wrong strategy.
“There’s been fairly substantial optimization of our marketing spend through this process,” he says, adding there are many other opportunities to improve marketing processes, including how he and his team recently deployed a new set of agents on top of these marketing datasets and data products, which can analyze performance data and make recommendations based on how the different campaigns are doing.
“Soon we’ll see agents autonomously making decisions and changing our campaign strategies in real-time,” he says. “Maybe there’s a human in the loop or maybe this runs fully autonomously.”
But the success of this initiative doesn’t mean Georgia-Pacific plans to throw AI at every business problem. “You must start with the business problem in mind, not AI, and think about it holistically,” he says. “To achieve what we have with our marketing team, we went on a three-year journey to get the data right and make sure what we were proposing was the right fit for them.”