categorySlug: finance category: Finance
Finance AI is an operating-model question
For a CFO, the difficult question is no longer whether finance should experiment with AI. It is how to make the work dependable enough to use across the organization. A promising demonstration is not a finance capability. Scaling it means confronting resistance to change, constrained budgets, legacy systems and talent gaps. [sourceRef: EY-WSJ-2026-10-28]
That puts AI on the finance leadership agenda, not just the technology shortlist. The task is to connect use cases to business priorities, reshape processes around them and build the data, skills and controls that support sustained adoption. EY’s account of finance AI is best read as an execution agenda—not a reason to chase every new tool. [sourceRef: EY-WSJ-2026-10-28]
Make sponsorship visible—and learning practical
Giving employees access to copilots does not, by itself, change how finance works. People need to know why a tool is being introduced, where it fits and what remains their responsibility. The CFO can set that direction, sponsor a deliberate change plan and give employees bounded, relevant tasks through which to learn. Change champions within finance teams can turn abstract guidance into practical habits and surface concerns early. [sourceRef: EY-WSJ-2026-10-28]
Draw the line between assistance and accountability. AI might accelerate analysis or help draft routine content; people remain responsible for financial decisions and for scrutinizing outputs. Clear guidance on appropriate use, review and escalation gives employees a sound basis for using the tools—not simply permission to try them. [sourceRef: EY-WSJ-2026-10-28]
Put pilots in a portfolio tied to business outcomes
A parade of disconnected demonstrations can generate activity without improving finance’s performance. Build a portfolio around the outcomes the business needs: perhaps better forecasting, a faster close, lower cost-to-serve or reduced working capital. Each initiative needs an owner, a route into day-to-day operations and a way to test whether it delivers the intended result. This is transformation work, not a collection of technology trials. [sourceRef: EY-WSJ-2026-10-28]
And finance cannot draw the boundary of that work at its own department. Consider procurement agents placing replenishment orders without human intervention. Finance may then need to revisit how commitments, approvals and monitoring work. Purchasing may sit elsewhere, but its consequences reach finance’s controls and records. Bring finance, technology, operations and HR into the work early enough to identify dependencies before they become operating problems. [sourceRef: EY-WSJ-2026-10-28]
Redesign the work before automating it
Put AI into a broken process and the process is still broken. The opportunity is to reconsider the journey end to end. In billing, that means looking beyond invoice production: ERP configuration, links to contract-management systems and customer-facing AI portals can all shape the experience. The design must account for handoffs, exceptions and controls—not only the step selected for automation. [sourceRef: EY-WSJ-2026-10-28]
Start with the work as it actually happens. Where do inputs come from? Who makes the decision? What happens when information is missing or a case falls outside the usual pattern? For a billing dispute, AI might draft replies to routine inquiries while a billing manager reviews and approves nonstandard cases. That division can speed routine work without assuming every case is routine. [sourceRef: EY-WSJ-2026-10-28]
Bring tax, internal audit and control owners into process design. Automation may change existing controls or create a need for new oversight. Specify who can authorize an action, what needs review, what evidence must be retained and where exceptions go. Revisit those arrangements when the process, model or business context changes, ensuring compliance and operational integrity are preserved end-to-end. [sourceRef: EY-WSJ-2026-10-28]
Make data readiness part of the business case
AI at scale depends on information that is reliable, governed and understood. Data stewardship is not a cleanup job to defer until after deployment; it is part of what makes deployment possible. [sourceRef: EY-WSJ-2026-10-28]
Large organizations may depend on several ERP platforms and other applications. Replacing them all with one system is not the only option: EY describes finance data appliances drawing information from more than 50 systems to support reconciliations and insights across business units. The architecture will vary, but the finance questions are consistent: Are terms defined consistently? Who owns the data? Can its lineage be followed? Who may access it, and how are discrepancies resolved? [sourceRef: EY-WSJ-2026-10-28]
Finance and technology should share responsibility for answering those questions. Decide how data is prepared for each use and how an AI-generated result can be traced to source records. Check the quality and coverage of the information before relying on generated analysis. Processing fragmented, poorly defined financial data does not make it trustworthy. [sourceRef: EY-WSJ-2026-10-28]
Match talent and funding to the work
A finance AI agenda needs people as well as tools. Assess the capabilities already on the team, identify reskilling needs, recruit selectively for data and technology expertise, and reconsider roles as the work changes. The point is talent strategy in service of a specific operating model, not training for its own sake. [sourceRef: EY-WSJ-2026-10-28]
Finance leaders also need to make resource needs visible. AI projects across the enterprise may compete for the same funding and technical capacity. Make a clear case for finance use cases, linking intended outcomes to the resources required. Decide what the team should build itself and where outside support makes sense. An AI center of excellence can spread methods and skills across functions rather than leave each department to assemble a separate capability. [sourceRef: EY-WSJ-2026-10-28]
Technical fluency matters, but so does the judgment that defines good finance work. Employees need to question assumptions, recognize anomalies and know when to escalate. Training tied to actual processes and controls, and repeated as those processes evolve, is more useful than a one-off event. [sourceRef: EY-WSJ-2026-10-28]
Use partners to build capability, not replace ownership
External providers can help finance leaders identify established use cases, accelerate implementation and develop internal capability. The test is what remains when the engagement ends: can the organization operate and govern the work better than before? The CFO still owns use-case selection, risk tolerances and fit with finance processes. A partner does not inherit that accountability. [sourceRef: EY-WSJ-2026-10-28]
Begin with work whose process and intended outcome can be described clearly. Set measures before implementation, test whether benefits persist beyond the initial trial and expand selectively. A use case that suits one organization may not suit another; data, process maturity and control requirements all matter. [sourceRef: EY-WSJ-2026-10-28]
Judge performance and control together
Speed and cost are only part of the scorecard. Measure the outcome the use case was meant to improve, then watch what happens in real work: how often people change or reject an output, what exceptions arise and how serious they are, and whether controls still work as intended. Compare results with a baseline and name who is responsible for responding when performance slips. [sourceRef: EY-WSJ-2026-10-28]
Expansion should follow evidence, not excitement. Define the use case and its boundaries; validate data and controls; test representative scenarios; review results; and widen deployment only when the evidence supports it. Continue monitoring changes in inputs, model behavior and workflow context. If automation changes who initiates or executes a transaction, reconsider where approval belongs. A compelling demonstration may earn a place in the portfolio; reliability earns a place in the operating model. [sourceRef: EY-WSJ-2026-10-28]
The CFO’s execution agenda
The advantage will not go to the finance team with the longest list of experiments. It will go to the team that connects a worthwhile use case to a redesigned process, trustworthy data, capable people and controls that hold up in practice. Sponsorship makes change possible; process design gives AI a useful role; governance makes that role possible to extend with confidence. [sourceRef: EY-WSJ-2026-10-28]
For CFOs, progress is not adoption for its own sake. It is the move from experimentation to repeatable value: better finance performance and insights for business partners that were previously difficult to obtain. [sourceRef: EY-WSJ-2026-10-28]
Source
Juan Uro, “CFO Take: 6 Actions to Scale AI in the Finance Function,” The Wall Street Journal CFO Journal, paid program created by EY, October 28, 2026. https://partners.wsj.com/ey/cfo/cfo-take-six-actions-to-scale-ai-in-the-finance-function/ [sourceRef: EY-WSJ-2026-10-28]