Enterprises Trade Big Data Cleanups For AI That Resolves Debt Incrementally

The most valuable asset an enterprise has for AI is the data it generates internally. It is unique, proprietary, and inaccessible to competitors. The problem is that the systems housing that data were never designed with AI in mind, and the procedural ownership around those systems lags even further behind.

Kjersten Moody is the CEO for North America at Elai, a platform that uses GenAI reasoning to autonomously generate predictive AI models from raw data. She previously served as the inaugural Chief Data Officer at Prudential Financial and the inaugural CDAO at State Farm, delivering AI solutions with up to 40x ROI across three Fortune 100 companies.

"AI needs a foundation. Without clean, well-owned data, your competitive advantage stops before it starts. The systems that generate that data were never built for AI, and that disconnect is where most companies stall out," says Moody. Data debt goes beyond technical implementation. It encompasses disconnected systems, siloed ownership, and poor quality relative to what AI requires. A customer name field that is blank half the time reflects a procedural ownership gap that no one is accountable for fixing. The older the system, the deeper the debt compounds.

• Brute force be gone: The traditional approach keeps legacy systems in place and builds routines to move data into a cleaned, enriched location. "This is where there is a crowded field of tools to help organizations move data into a central environment, and these projects are frequently expensive and slow," says Moody. Ripping out legacy infrastructure entirely carries enormous risk with no guarantee of a smoother outcome.

• Agentic AI changes the math: A new generation of solutions can take hundreds of data sources and reconcile, enrich, and cleanse them automatically. This decouples data debt from AI value creation, letting enterprises generate ROI while deciding how to reinvest. "You can start to generate benefit from AI, which then gives you discretion as a management team," Moody says. "Do we want to take some of that benefit and use it as found money to invest in the foundation building?"

Moody argues the shift also demands a different organizational approach. Top-down mandates are losing ground to bottom-up, process-level transformation.

• Start at the process level: Rather than engineering one strategy for the entire enterprise, Moody recommends mapping transformation plans at the process level, then identifying common denominators. "Give as much autonomy as possible to those process definitions and process owners, with a lighter touch coordination and enablement from supporting teams."

• Context engineering is non-negotiable: AI requires the right data in the right context at the right moment. "Context engineering is a new dimension, but at the end of the day, you're engineering quality and trusted solutions," says Moody. "You have to know when you're not ready so you're not overextending into something that isn't going to work properly."

• Data retrofits are messy: Data retrofits are messy. Moody warns teams to prepare for data that is significantly more imperfect than expected. "We're going to kick over a lot of rocks, and we have to expect that some of those rocks have snakes underneath. It's about being very pragmatic about what it's going to take to move forward and selecting the right tools to help navigate and accelerate ROI delivery."

The contrast with AI-native companies makes the challenge sharper. Organizations born in the AI era design processes by asking what computers should do first and then filling in the rest with people. Legacy enterprises carry the opposite bias, having built processes around human workflows and layered technology on top. Incumbents must untangle and rebuild what AI-native competitors never had to assemble, and the data layer is where that work begins.

"AI-native companies don't have the history of the old way of designing process," Moody concludes. "They have a different perspective, and for the young AI-native companies, that has been the only perspective, the dominant perspective."