Articles
December 17, 2025
AI Advances Fast, but Treasury Still Needs Better Data Plumbing
# Technology
# Sponsored Content
Treasury4 says clean, connected data is the prerequisite for AI to move from promise to daily impact.
AI is beginning to reshape treasury teams’ daily work, but the biggest gains still depend on data arriving clean, complete and on time. That theme ran through a NeuGroup interactive session with Treasury4, now strengthened by its acquisition of TreasuryGo and the addition of its co-founder—and longtime Microsoft treasurer—George Zinn as president.
- Alongside Mr. Zinn, Treasury4 co-founder Ed Barrie and chief technology officer Greg Morris stressed that AI cannot overcome timing gaps, inconsistent reporting or fragmented data—problems that increasingly require treasurers to focus first on data plumbing before full implementation.
- Mr. Zinn recalled working with Mr. Barrie in treasury at Microsoft during the 2008 financial crisis, collaborating to create a real-time, aggregated view of cash across the company’s banks—a challenge made difficult by systems that didn’t align, statements that didn’t match and banks using different FX rates. “What I love about Treasury4 is we’re now building the tools we wish we had at the time,” Mr. Zinn said.
George Zinn, Treasury4
Hiring a plumber. Mr. Morris said not all treasury teams have the bandwidth to build or maintain the data pipelines that modern forecasting and automation require. The real work, he said, is ensuring that the underlying infrastructure reliably connects banks, ERPs and internal systems so data flows into the platform in a consistent, usable form.
- He added that while a few large companies build this infrastructure themselves, most need a service that centralizes and maintains those connections as institutions and systems evolve. Treasury4’s work is “accessing the right data, organizing that data, categorizing that data, and then operating on that data.”
- Once that foundation is in place, companies can choose how deeply they want to engage with AI; “whether you do that via your own tools or use a service-provided tool is entirely a question of how deep you want to get into AI.”
Greg Morris, Treasury4
The foundational problem: inconsistent data flows. During the session, members were asked which area of treasury would benefit most from wading into the AI pool. More than half (53%) chose cash flow forecasting, far ahead of reconciliation at 31%. But for many corporates, the data that would feed those forecasts often arrives out of sync, incomplete or in formats that don’t line up. As Mr. Zinn noted, trying to forecast without a stable starting point “is like getting directions when you don’t know where you are on the map.”
- Members described manually updated balances, mismatched statement dates, cutoff differences across time zones and cash in transit as sources of noise that weaken forecasts. One said these issues can “cast doubt on the integrity of the model” before any analysis even begins.
- Mr. Barrie said teams should capture as much detail as possible at the transaction level, as part of initial plumbing, because future use cases often depend on fields that don’t seem essential today. “The joke internally is that there’s no data point Ed won’t want mapped into the system,” he said, emphasizing that clear definitions of what each field means—and how data should be used—directly affects how well AI tools perform.
- Members asked whether AI could help explain exceptions—for example, discrepancies between bank-reported balances and what shows up in internal systems. Today, teams often must investigate these gaps manually. Mr. Barrie said Treasury4 is building logic to identify root causes such as missing transactions, reporting delays or categorization issues, reducing the need for manual fixes.
Ed Barrie, Treasury4
Putting the pipes to work. With consistent data structures and stable connectivity, Treasury4 applies agentic AI to accelerate visibility and decision-making. “By tagging the cash flows,” Mr. Zinn said, “we’re able to back into a direct cash flow statement that allows us to make a go-forward forecast within the system and basically stress-test different scenarios.” Cleaner feeds and richer bank APIs move treasury closer to real-time liquidity views.
- Mr. Morris said the team is adapting to a world where liquidity moves faster and across more channels, from money market portals to digital settlement rails. Treasury4 is using AI internally to shrink development cycles, compressing work that once took months into hours and supporting more intuitive user interfaces. Mr. Barrie added that AI is now linking bank and ERP data to validate beneficiaries, detect anomalies and improve routing accuracy as fraudsters use more advanced tools.
- With the plumbing in place, AI can scale work treasury teams have long struggled to maintain manually. Without it, even the strongest models remain constrained by data that’s incomplete or doesn’t line up across systems.
Sponsored by:
Dive in