Half of Finance Teams Are Stuck in the Middle on AI Maturity
AI in finance isn't early anymore-it's uneven. A global survey of 1,520 finance and business leaders from Payhawk shows half of organisations now sit in the "middle": actively experimenting with AI but struggling to scale it safely into core workflows. As CFOs enter budget season, that middle is where execution risk piles up.
The middle is the market's centre of gravity
On a 1-10 maturity scale, around 50% of organisations rate themselves 4-6. They have activity, pilots, and pockets of value-but AI isn't a core finance capability yet. Nearly one-third call themselves 7-10, which makes "leader" a broad label that hides very different operating realities.
The gap matters more in finance than most functions. AI must pass controls, audit, accountability, and policy enforcement before it touches workflows that move money. As Payhawk's CEO, Hristo Borisov, puts it: "The real risk in finance AI isn't experimentation, it's getting stuck halfway." The teams that win won't be the loudest-theyll be the ones that make AI governable inside the finance operating model.
Context drives maturity
Maturity varies sharply by industry and size. Tech organisations with more than 251 employees show the highest maturity globally, with over 70% self-rating as highly mature. Smaller companies in regulated and core-economy sectors (50-250 employees) report just 13.5% in the high band. Large non-tech firms mostly sit in the middle-adopting, but not scaling into core operations.
There's a structural thread behind this. Multi-entity organisations tend to rate higher because scale forces standardisation, shared services, and centralised controls. But there's a catch: without consistent data and alignment across entities, governance slows down and progress stalls.
"Leaders" aren't one thing
Self-identified leaders break into very different camps. Some have embedded AI into defined workflows with clear accountability. Others are moving fast without minimum guardrails-or investing heavily without the foundations to scale. The constraint isn't model capability; it's whether adoption can be made stable, defensible, and repeatable inside financial control environments.
How CFOs move from experiments to scaled, auditable AI
If you sit in the middle, treat scale as a control problem first, a model problem second:
- Use-case inventory: Map high-value finance use cases (close, AP, T&E, forecasting, reconciliations) to objectives, materiality, and data needs. Kill nice-to-have pilots.
- Policy and controls: Stand up an AI policy that covers approvals, data use, model inventory, human-in-the-loop checkpoints, and separation of duties.
- Data foundation: Standardise chart of accounts, vendor master, entity mappings, and reference data. Add lineage and quality SLAs so models aren't guessing.
- Auditability by design: Log prompts, responses, versions, training data sources, and decisions. Keep reproducible records that survive audit.
- Risk management: Map controls to frameworks like the NIST AI Risk Management Framework. Cover privacy, security, model drift, bias checks, and regulatory needs (e.g., SOX, GDPR).
- Vendor diligence: Require SOC 2, DPAs, sub-processor transparency, and clear exit paths. Score vendors on finance-grade controls, not just features.
- Integration first: Connect AI to ERP, AP, expense, and close systems via supported APIs. Prefer native connectors over fragile workarounds.
- Operating model: Assign product owners, set RACI, and schedule model/controls reviews. Add FinOps to track usage and unit economics.
- Measurable scale: Define KPIs-days to close, exception rates, cost per invoice, forecast accuracy, audit findings-then gate progression from PoC to production.
- People and training: Upskill controllers, FP&A, and AP teams on prompts, review protocols, and exceptions. Centralise enablement to avoid one-off methods. For curated training options, see AI Learning Path for CFOs and the AI Learning Path for Finance Managers.
Budget moves for the next cycle
- Prioritise data standardisation and entity alignment over net-new pilots.
- Fund control tooling: access controls, audit trails, model versioning, and monitoring.
- Pay down integration debt in ERP/AP/expense/close. That's where scale lives.
- Ring-fence budget for change management and training. Adoption is a management problem.
The takeaway for finance leaders
Half the market is stuck in the middle. The way out is boring on purpose: standards, data, controls, and repeatable integrations. Get those right, and AI can move from activity to operations without blowing up audit or policy.
The first installment of Payhawk's CFO AI Readiness Report is available on the company's site.
Your membership also unlocks:
