AI scaling stalls as financial leaders face governance and data gaps

Written by Pooja Sharma

Deep Dive” We explore relevant topics, hot trends, and new products every two weeks. Exclusive to Prime members.

How does it work?

New research from Payhawk reveals that ‘AI leaders’ are divided into six operational postures rather than a single maturity path

Nearly half (45%) of organizations that consider themselves “AI leaders” lack the baseline governance needed to securely scale AI in their finance workflows, according to new research from Payhawk.

This study also challenges the common assumption that AI maturity progresses along a defined maturity ladder. Even within the “Leader” category, AI readiness is divided into distinct adoption postures, each constrained by different readiness gaps. The data reveals that the real constraint to scaling is not the power of AI, but governance – the ability of organizations to defend, track, and audit what AI does within financial workflows.

These findings are based on a global survey of 1,520 financial and business leaders. “AI Leaders” (subset n=405) are defined as organizations that rate their AI maturity between 7 and 10 out of 10.

“Rules” and “data” debt: Why AI programs stall

Five operational requirements determine whether AI can move from “deployed” to “operational” within finance workflows. they are execution Minimum measures have been taken rule For AI use, skill and tools, dedicated budget and data Can be used for AI analysis. Only 26% of AI leaders meet all five requirements.

The study categorizes leaders into six operational postures based on their scores on five requirements:

• Extended adopters (26.9%)— Strong in all five requirements. These organizations have a complete operational stack.
• Progressive improver (17.5%)— There are several AI readies across the stack, but no one aspect is decisively powerful.
• Execution-driven implementers (16.0%)— Good execution and skill, but minimal rules. This is the most obvious “rules debt” attitude.
• Agent first, control second (14.1%)— Enthusiasm and experimentation trump governance. There are no minimal rules and there is limited readiness for execution.
• Scalers who value governance (13.8%)— Strong rules and enforcement, but weak data preparation (only 30% strongly agree). This is the most obvious “data debt” attitude.
• Control-first planner (11.6%)— Skills, budget, and data are relatively strong, but execution is not in place. Features exist without deployment.

Who is under the radar?

These two systemic gaps, “rule debt” and “data debt,” explain why scaling fails.

• rules debtThis happens when organizations deploy AI before establishing governance, leading to systems that cannot be audited, accounted for, or securely incorporated into workflows around approvals, compliance, and financial management. Execution-driven implementers and agent-first, control-second attitudes exhibit this pattern, and together they make up about 30% of leaders.
• data debt It occurs when governance and execution are in place, but the underlying data is inconsistent, incomplete, or fragmented. In these cases, organizations can control the use of AI, but cannot trust its output at scale. The most obvious career is the scaler, who drives governance.

Why is it important now?

This study highlights a clear imbalance. Although 78% of self-proclaimed “AI leaders” report having strong skills and tools, only 55% have minimum governance rules in place, ranking them among the lowest for readiness factors.

Rule debt explains why many organizations appear to be “advanced” in their activities but still struggle to move beyond narrow support use cases. This often focuses on smaller, faster-moving situations. Data debt, on the other hand, explains why some organizations, even if they appear disciplined and well-managed, are unable to scale AI into their core financial operations. Data debt is concentrated in complex and controlled contexts. This study also identifies common and costly mismatches. Organizations invest more in AI capabilities when the real hindrance is governance infrastructure, and build policy frameworks when the real hindrance is data quality. In both cases, progress stalls because the operational constraints being addressed are not the only limiting scale.

“Scaling financial AI feels inconsistent as organizations progress unevenly across the capabilities that support scale.” Hristo Borisov, CEO and co-founder of Payhawk, said: “Many organizations are making further investments in AI even though the real bottlenecks lie elsewhere: in rules and data. Scaling AI in finance is fundamentally an orchestration challenge: coordinating rules, data, and accountability across workflows. Organizations that only address some readiness requirements face unique trade-offs and end up stuck with supportive use cases.”