AI reduced the work, but human oversight became the bottleneck.

AI is making finance teams faster. The results aren't showing up where it counts – and the reason has nothing to do with the tools.

The tool worked. Reconciliation that took three days now takes three hours. AP and AR responses that once required a dedicated analyst are drafted in seconds. By any reasonable measure, the finance team is faster.

And yet the close takes the same number of days. The CFO is still waiting on the same sign-offs. Audit preparation still hits the same wall two weeks before quarter-end. This is the underreported reality of AI in finance right now. Not failure. Not resistance. The constraint moved.

The bottleneck is no longer the work itself. It's everything that happens after the work is produced. Data still needs to move between systems and exceptions still need to be reviewed. Decisions that come from this still require extra approvals. Finance operating models were designed for a world where humans generated outputs, and AI is exposing the limits of that design.

We recently commissioned Forrester Consulting to survey 1,279 finance decision-makers across EMEA, APAC, and North America (NA). The findings reveal a gap between how fast finance teams are moving and how ready their systems are to keep up.

All statistics referenced below come from "Building An AI-Ready Finance Function", a commissioned study conducted by Forrester Consulting on behalf of Airwallex, June 2026.

Finance is getting faster, the system needs to catch up

One in three finance leaders describe their workflows as fully digital. But 84% still require manual intervention to complete those same workflows. This sounds contradictory until you look at how work actually moves across finance teams.

In many companies, "fully digital" simply means work happens inside software rather than on paper. Data may still be moving between disconnected systems, and that means teams still manually reconcile information across functions. Reviews, approvals, and exceptions still depend on people coordinating across workflows that were never designed to talk to each other.

The data fragmentation picture sharpens this further. 65% cite fragmented or inconsistent finance data as the top technology barrier to scaling AI. 84% still require manual work to complete finance workflows. And the gap is uneven across functions, running from 41% in tax and treasury up to 72% in corporate development and investor relations, which are precisely the functions where strategic decisions get made.

Having a single core platform doesn't always solve this problem. While 30% of organisations run on a single ERP, only 16% report consistent data flow. This matters because fragmented data creates more review points, more reconciliations, and more opportunities for human intervention. In other words, it creates new bottlenecks precisely where AI is trying to remove them.

AI automates one task, and the bottleneck moves to the next one

That unevenness matters, because finance doesn't operate as a collection of individual tasks. It operates as a system. A workflow is less like a series of isolated activities and more like a relay race. The team only finishes as fast as its slowest handoff. If one runner suddenly gets much faster, the baton may reach the next runner sooner, but the overall result is still constrained by the next handoff. In finance, AI often accelerates individual tasks, while approvals, reviews, and decisions remain the limiting factor.

AI can draft reports, reconcile transactions, and surface anomalies faster than before. But the outputs still need to be reviewed, approved, explained, and incorporated into broader decision-making processes. The analyst who spent three days on reconciliation now has those days back. But the output still sits in a queue waiting for the controller, the auditor, and the CFO.

The sign-off chain doesn't compress because the upstream work got faster. The reviewer becomes the bottleneck, who then determines the speed of the whole system. That’s why many organisations see strong productivity improvements without seeing equivalent improvements in cycle times.

Accountability doesn’t move at machine speed

The bottleneck exists because finance operates under legitimate accountability requirements. The top challenges finance leaders cite when scaling AI reflect this reality. Risk and compliance concerns (44%), limited trust in AI outputs (38%), and unclear expectations around AI's role (34%) consistently rank among the top constraints. These are governance problems, embedded in how finance operates for good reasons.

Finance sits at the intersection of regulatory compliance, financial reporting, risk management, and executive decision-making. Every output carries accountability. Someone must be able to explain it, defend it, and sign their name to it. That responsibility doesn't disappear simply because AI generated the output faster.

This pressure is intensifying. Growing economic uncertainty is compressing decision timelines at the same time when governance requirements are expanding. 75% of finance leaders rank strengthening financial controls, audit readiness, and regulatory compliance as the top strategic priorities this year, alongside AI integration and enterprise-wide resilience against market volatility. Regulators and auditors are moving toward greater scrutiny of AI-generated outputs, a sign that human oversight requirements are being formalised.

This means finance teams face a new challenge. The question is no longer how to automate work. It's how to maintain accountability when work can be produced at machine speed.

Automation solves tasks, finance needs to solve flow

That tension between machine-speed execution and human-speed oversight points to a broader concern. If AI accelerates individual tasks but work still stalls at reviews, approvals, or handoffs, the issue is no longer automation. It's how work moves across systems, teams, and controls.

For decades, finance operating models were built around a simple assumption: humans create and validate the work. The same teams responsible for producing reconciliations, forecasts, analyses, and reports were also responsible for validating them. Governance was built around overseeing human-generated output. The pace of decision-making naturally matched the pace of execution.

AI changes that equation. Analysis, anomaly detection, reconciliations, and recommendations can now be generated before a finance professional becomes involved. The role of finance is gradually shifting from producing information to supervising systems that produce it.

But governance often remains calibrated to a much slower operating model. As a result, many organisations find themselves with machine-speed execution sitting behind human-speed oversight. The challenge is no longer creating information. It's deciding what requires human review, what can be automated, and how decisions move through the organisation with appropriate controls.

This is why many finance transformation efforts are moving beyond task-level automation and toward orchestration. The focus is increasingly on creating connected workflows where data moves consistently across finance systems, operational platforms, banking infrastructure, and reporting environments. AI-generated outputs carry context, approvals, and audit trails with them. Exceptions are automatically routed to the right reviewer. Decisions, approvals, and audit trails travel with the workflow rather than being recreated at every stage.

The goal is to reserve human attention for where judgement adds the most value. Instead of reviewing every transaction, teams focus on material exceptions. Instead of manually consolidating information across systems, they work from a shared view of the data. Processes such as the close, audit preparation, and forecasting become more continuous rather than repeatedly bottlenecking at handoffs and sign-offs.

The real shift: from human-generated work to human-supervised work

The next phase of finance transformation won't be defined by who deploys the most AI tools. It will be defined by who redesigns accountability, oversight, and decision-making for an environment where execution happens continuously.

The teams that succeed won't necessarily have more AI, because the answer isn't more AI. It's redesigning the system around what AI has already changed. It’s about having operating models capable of turning AI-generated outputs into decisions, actions, and outcomes without creating new bottlenecks along the way.

In practice, that means reducing the friction that accumulates between systems, teams, and controls. Data needs to move consistently across finance platforms, operational systems, banking infrastructure, and reporting environments. AI-generated outputs need to carry the context, approvals, and audit trails required to move through governance processes without being recreated at every stage. Human attention needs to focus on material exceptions and judgement calls, rather than reviewing every transaction by default. The close, the audit, and the forecast become continuous processes instead of discrete ones that bottleneck at sign-off.

The infrastructure to support this exists. The question is whether finance teams are prepared to redesign operating models that were built for human-speed execution, now that work can be generated at machine speed? That may ultimately determine which organisations translate AI-driven productivity gains into faster decisions, stronger controls, and better business outcomes.

Source: Unless stated, all statistics referenced in this article come from "Building An AI-Ready Finance Function", a commissioned study conducted by Forrester Consulting on behalf of Airwallex, June 2026.