AI for CFOs
A weekly series for CFOs and finance leaders on applying AI to the high-volume, rules-heavy work in finance: purchase-to-pay, order-to-cash, the close, compliance and multi-entity reporting. What works, what still needs a person, and how to keep control while the routine moves to software.
AI in the Office of the CFO: A Practical Guide for 2027
Most finance teams now use some AI, and few can point to a large result from it. This guide sets out where AI earns its place in the CFO's office, what still needs a person, which controls to put in first, and a realistic plan for the next twelve months.
Everything we've published.
26 insights
AI in the Office of the CFO: A Practical Guide for 2027
Most finance teams now use some AI, and few can point to a large result from it. This guide sets out where AI earns its place in the CFO's office, what still needs a person, which controls to put in first, and a realistic plan for the next twelve months.
SAP ECC End of Maintenance: What 2027 Means for Finance Teams
SAP's mainstream maintenance for ECC ends with 2027, and extended maintenance to 2030 costs more, so ECC companies face a migration on a timetable set by the vendor rather than the close calendar. This piece sets out the dates, what they mean for finance, and how to keep controls and exception management steady while the ERP underneath changes.
E-Invoicing Mandates 2026–2028: A Timeline for Multi-Country Finance Teams
E-invoicing rules now change an invoice's format, route and timing country by country, and several of the next deadlines fall on 1 January 2027. This guide sets out what is already in effect, what is coming to 2028 and beyond, and what a multi-country finance team should do in the weeks it has left.
Measuring AI ROI in Finance: Hours, Errors, Cycle Time and Cash
Boards are starting to ask what this year's AI spend in finance returned, and a vendor dashboard of hours saved does not answer the question. This piece sets out four measures that do, how to count the cost side honestly, and how to read survey figures that seem to disagree.
AI Proof of Concept in Finance: Designing a Pilot That Reaches Production
Many finance AI pilots end with a good demonstration and no decision, because nobody agreed in advance what success meant or who would own the result. This guide sets out how to scope a proof of concept around one process, with read-only data, a measured baseline, agreed success criteria, a named control owner and a fixed decision date.
AI Variance Analysis: Commentary a Board Can Trust
AI can draft variance commentary quickly, and a wrong explanation reads exactly as confidently as a right one. This piece sets out the four things that make AI-drafted commentary safe to put in front of a board: figures from the ledger, explanations checked against evidence, agreed materiality thresholds and a named reviewer.
AI in the Month-End Close: Reconciliations, Flux Analysis and Commentary
AI can now match transactions, explain reconciliation differences and draft close commentary, which is most of the work that keeps a close running late. The question for a CFO is which of those outputs can be relied on as they come, which need a reviewer, and what controls keep the close auditable once AI is part of it.
AI in Order-to-Cash on Microsoft Dynamics 365: Where to Start
Dynamics 365 Finance now puts AI inside the customer record and the reconciliation workspace, yet the question a CFO asks each week, which customers need a person and who is on it, still spans hundreds of records. This piece sets out what Microsoft provides, what a read-only layer over Dynamics adds, and a sensible order in which to start.
AI in Accounts Payable on SAP: What Works and What Still Needs People
AP is where finance teams most often start with AI, yet most of the effort on an SAP payables team goes into exceptions, disputes and approvals that carry authority. This piece separates what AI does reliably in accounts payable today from what still needs a person, and compares adding a read-only layer over SAP with changing SAP itself.
AI Internal Controls: What Changes When AI Touches Financial Reporting
AI is moving into reconciliations, journal suggestions and commentary, which means it is moving inside the controls that auditors test. This guide sets out what changes in control design, evidence, change management and segregation of duties, and what to raise with your auditors before year-end rather than during it.
An AI Use Policy for the Finance Function: What to Put in It
Most finance teams now use AI in some form, and many could not show an auditor or an audit committee the rules they follow when they do. This guide sets out what a finance AI use policy should contain, with a section-by-section outline, and how it lines up with ISO/IEC 42001 and the EU AI Act as they stand in August 2026.
AI Guardrails for Finance Data: Access, Logging and Human Sign-Off
Finance holds the data a company can least afford to leak or misuse, and AI tools make it very easy to move that data somewhere new. This piece sets out six practical guardrails for finance data and how they line up with the NIST AI risk frameworks.
AI With Company Context: The Layer Most Rollouts Skip
A general AI assistant can explain how approval limits usually work, but it cannot tell an AP clerk what your own delegation of authority says. This piece covers the company context AI needs in finance and how to supply it without losing control of who sees what.
Enterprise AI Adoption: Why Giving Everyone a Chatbot Is Not a Strategy
Most companies have now given their staff a general AI assistant, and few CFOs can name a process that runs differently because of it. This piece sets out what enterprise AI adoption needs beyond licences, and why finance so often ends up holding it.
A Compliance Calendar for Multi-Country Groups: Why Spreadsheets Fail at Ten Jurisdictions
A statutory compliance calendar kept in a spreadsheet works for two or three countries and quietly stops working somewhere past that, usually without anyone noticing until a deadline is missed. This guide sets out what replaces it: an obligations register, named owners with maker and checker, evidence, escalation, and a routine for catching rule changes.
Financial and Operational KPIs in One View: What the ERP Cannot Give You
The numbers that explain the P&L sit in the CRM, the payroll system and operations tools, not in the ERP, and each team brings its own version to the monthly review. This piece covers how to combine them, the definitions and ownership work that has to come first, and a short KPI list by business model.
Multi-Entity Reporting Across Different ERPs, Without a Data Warehouse Project
Most groups that grow by acquisition end up with entities on different ERPs, and the consolidated pack is rebuilt by hand in a workbook every month. This guide sets out a mapping layer, a common chart, consolidation and eliminations, and governance that give a reliable group view without waiting for one ERP or a data warehouse.
Intercompany Reconciliation: Why It Breaks Across Entities and Systems
Intercompany balances should cancel to zero on consolidation, yet in many groups they are the last thing holding up the close. This piece sets out why the two sides disagree, the matching rules and settlement routine that keep them aligned, and where AI matching suggestions help.
Collections Prioritization: Who to Call First, and Why the Aging Report Gets It Wrong
Most collections teams work the aging report from the oldest bucket down, which feels disciplined and often sends the best effort to the least collectible balances. This piece sets out what to rank on instead and a simple daily worklist method, with the parts AI does reliably and the parts that stay with a person.
Unapplied Cash and Short Payments: The Quiet Drag on DSO
The customer has paid, the money is in the bank, and the invoice still shows as overdue. This piece explains why cash goes unapplied, what it does to DSO and to the collections list, and how to own it, age it and clear it, including where AI-assisted matching helps and where it does not.
Vendor Master Data Controls: The Weak Point Behind Payment Fraud
Invoice controls check what is being paid, while the vendor master decides where the money goes, and a single unverified bank-detail change can route a correct payment to the wrong account. This piece sets out the vendor master data controls that close that gap, and what AI can watch continuously between reviews.
Duplicate Payments: How They Happen and the Controls That Catch Them
Every ERP has a duplicate invoice check, and duplicate payments still go out, because the second invoice rarely looks exactly like the first. This piece explains how duplicates get through, which preventive and detective controls catch them, and where fuzzy matching and AI fit before the payment run.
Three-Way Match Exceptions: Why They Pile Up and What Clears Them
Many invoices on payment block are not vendor errors but timing, tolerance and data problems that land in a queue AP can see and cannot fix. This piece sets out why three-way match exceptions accumulate, who should own each type, and how to clear them with AI doing the sorting and people making the calls.
Management by Exception in Finance: Designing Work Around What Needs a Person
Most finance processes still ask people to look at everything in order to find the few items that need them. This piece covers how to design work so people see only what needs judgment, with thresholds, named owners, aging and escalation, and where AI helps and where it should stop.
AI Use Cases in Finance: How to Rank Them by Volume, Rules and Risk
Every finance team now has a long list of things AI could do, and the list is usually ordered by whatever looked best in a demo. This guide gives a five-factor scoring method, worked examples across payables, receivables, the close, reporting and compliance, and a way to turn the scores into a first project and a level of human review.
Finance Team Capacity: A Time Audit of Where the Hours Go
Most finance leaders know the team is stretched, and few can show with numbers where the hours go. This piece sets out a four-week time audit by process, volume and rework, and how to read it for the work that AI and better process should take off people's desks.