← Knowledge Base 🕐 11 min read

Knowledge Base

CFO AI Workflows

· March 2026

The specific finance function workflows where AI is producing documented results in 2025–2026, ranked by evidence qua...

The specific finance function workflows where AI is producing documented results in 2025–2026, ranked by evidence quality and implementation readiness. As of March 2026, only 17% of finance teams actively use AI in core workflows — finance ranks last among all business functions in deployment maturity (CFO Connect, March 2026). The gap is execution, not awareness: 87% of CFOs expect AI to be very or extremely important to their finance operations in 2026 (Deloitte Q4 2025 CFO Signals, n=200).

The Five Workflows, Ranked by Evidence Quality

1. Financial Close

The strongest independent evidence in this stack. MIT Sloan and Stanford researchers (Choi & Xie, August 2025, n=277 accountants, 79 SMBs) documented:

7.5 days reduced from the monthly close cycle
55% more clients served per accountant per week
8.5% shift from routine back-office to higher-value work
12% increase in financial report detail

The study used an AI accounting software provider as a partner but is peer-reviewed with named researchers — the only such dataset on close automation as of April 2026. The sample is SMB-weighted; large-enterprise ERP environments may see different results. Vendor claims (40–80% reconciliation workload reduction from BlackLine, Workiva) exist but are uncontrolled.

The Consero Global 2026 CFO Survey (n=102, $20M–$500M investor-backed companies, May 2026) adds a time-threshold benchmark for the mid-market: in 2024, 8% of finance teams closed the month in under 10 days; by 2026, 65% do. The eightfold shift happened in two years among companies with $20M–$500M revenue — not at large enterprise. Sample is vendor-sponsored (Consero is an AI-enabled finance outsourcing firm) and likely skews toward technology-forward operators; treat adoption percentages conservatively. The close-time threshold is the most specific mid-market benchmark available. Source: research/07-adoption-challenges/consero-cfo-survey-ai-finance-2026.md · May 2026 · MEDIUM / TIER 1

Best starting point: Reconciliation matching and intercompany elimination — high volume, low judgment, measurable.

2. AP/AR Automation

Strong vendor case study evidence with at least one independent data point (Fanatics Betting & Gaming, cited in Bain Capital Ventures CFO survey, n=50, Feb 2025):

AP month-end workflow: 20 hours → 2 hours
L.E.K. Consulting (n=~100 CFOs, Dec 2025): AP invoice processing 3 hours → 15 minutes

Vendor case studies (SoftCo/Logitech: 83% straight-through processing; Superdry: 5% to 80% touchless; Primark: 98% invoice match rate) are directionally consistent but uncontrolled. These case studies are vendor-published and represent selected wins with no control group and no independent verification.

The hardest problem is the 17–45% of invoices still requiring human exception handling. Automation investments should be evaluated on exception management capability, not straight-through rate alone.

3. FP&A and Variance Analysis

Adoption lags the opportunity. Accounting Today’s 2025 FP&A survey found:

Commentary/report writing: 57% AI adoption (most common use)
Variance analysis: 30% AI adoption
Forecasting and planning: 28% AI adoption

The two highest-analytical-value workflows are the least deployed. AI-generated variance narrative — explaining budget-vs-actual gaps from structured data — is technically well-suited to current language models. First-draft commentary generation allows the FP&A analyst to shift from writing to reviewing, reducing cycle time without reducing accountability.

PwC (2026) cites 30% reduction in report production time and 20% improvement in forecast accuracy from predictive analytics. Neither figure includes disclosed sample size or methodology — treat as directional.

4. Board Reporting

The most deployed use case by adoption count (57% of AI-using finance teams report using it for board reporting per CFO Connect 2026), but with the shallowest analytical depth. Current usage is primarily slide assembly, commentary drafting, and formatting — document production AI, not analytical AI.

The governance requirement here is non-negotiable: any AI-generated figure in a board package requires a defined human review step before distribution. The reputational and audit exposure from a transposed number is too high for experimentation without explicit accountability.

5. Covenant Monitoring

The least evidenced CFO AI workflow. No independent study provides time-reduction data. Vendors (Datagrid, CovenantIQ, Cardo AI, Moody’s Lending Suite) describe systems that shift monitoring from quarterly to daily, with predictive alerts weeks or months before potential breaches.

The business case is risk asymmetry, not efficiency: covenant violation costs (amendment fees, waiver costs, relationship damage, potential acceleration) are large relative to monitoring tool costs. For any company with active credit facilities carrying financial maintenance covenants, the ROI frame is risk reduction, not hours saved.

Adoption Pattern

L.E.K. Consulting’s 2025 OCFO survey (n=~100 CFOs) found:

11% actively use AI in finance functions
35% running pilots or proofs of concept
25% using AI-powered features embedded in existing third-party platforms
~56% prefer embedded AI over point solutions

The preference for embedded AI reflects the execution risk dynamic: CFOs want AI that integrates with existing ERP, FP&A, and close management systems rather than requiring new infrastructure.

Practitioner voices (pillar 13)

“We’re seeing that and 90% more in full… they also report that they’re saving time. They’re reducing manual work by 30%. 62% of them say that bookkeeping is easier.”

Mariana Tessel, Executive Vice President and General Manager, Intuit — Beyond the Pilot / VentureBeat, April 2026 · Source: research/13-multimodal-sources/beyond-the-pilot/2026-04-01-100m-agents-scaling-the-new-execution-stack-with-intuit.md

Relevance: Intuit Intelligence is the largest deployed AI accounting system for SMBs (3 million customers). Tessel’s 30% manual-work reduction and 62% bookkeeping ease figures are customer-reported outcomes at scale — directional but denominator-clear, which places them above typical vendor case studies. The pattern (time savings first, accuracy improvement second) matches the sequencing evidence from the MIT Sloan / Stanford close-cycle study.

“We try to ask the business… we try the commerce people or supply chain people or finance people, where do you think the big value could be?”

Ronald den Elzen, Chief Technology and Digital Officer, Heineken Group — Me, Myself, and AI / MIT SMR + BCG, January 2025 · Source: research/13-multimodal-sources/me-myself-and-ai/2025-01-07-how-a-160-year-old-startup-uses-ai-the-heineken-companys-ron.md

Relevance: Heineken’s CTDO describes the workflow-identification process as finance-inclusive from the start. This is the pre-condition the CFO AI deployment sequence requires: value identification is a joint exercise between digital/tech leadership and the finance function, not a top-down mandate. The finance people at the table define the prioritization; the technology follows.

“We have invested more than $3.5 billion in data and AI infrastructure alone for the past 10 years.”

Sam Hamilton, Senior Vice President of Data and AI, Visa — Beyond the Pilot / VentureBeat, October 2025 · Source: research/13-multimodal-sources/beyond-the-pilot/2025-10-06-venturebeat-in-conversation-visas-35b-bet-on-ai.md

Relevance: Visa’s $3.5B infrastructure investment over 10 years is the upper end of the financial-services AI investment arc, relevant to CFOs as a benchmark for what “serious AI in finance” looks like at scale. The figure contextualizes mid-market investment decisions: Visa’s investment is not a target, it is evidence that durable AI value in finance requires multi-year capital commitment, not annual tool purchasing.

“The number one fear that we hear from finance leaders is, ‘I don’t trust it.’ […] Seventy-one percent of finance leaders say they would reject AI that can’t explain itself.”

Jeremiah Edwards, Head of Sage AI — Everyday AI Ep 772 / Jordan Wilson, May 2026 · Source: research/13-multimodal-sources/everyday-ai/2026-05-22-ep755-780-enterprise-ai-mining.md

Relevance: The PwC partnership study Edwards cites (71% of finance leaders rejecting non-explainable AI) is the clearest single-number expression of the trust floor in finance AI deployment. Every AI workflow in this page that touches external reporting, audit, or regulatory filings requires a full reasoning trail — not as a nice-to-have but as the explicit acceptance criterion for 71% of the buyer population. Note: Sage is a vendor; product performance claims carry LOW credibility. The PwC 71% stat and the governance framework described are independently citable. Source: Everyday AI Ep 772 “AI You Can Trust,” May 7, 2026, guest Jeremiah Edwards, Head of Sage AI.

Evidence Base

88% of organizations claim AI adoption; only 6% generate meaningful profit from it — McKinsey (cited in Everyday AI Ep 755, April 2026, and Ep 780). The gap is not access to better models; it is change management and workflow redesign. For CFOs, this is the accountability problem: investment is nearly universal, measurable return is not. research/13-multimodal-sources/everyday-ai/2026-05-22-ep755-780-enterprise-ai-mining.md
Only 17% of finance teams actively use AI in core workflows — CFO Connect (March 2026). Finance ranks last among all business functions in deployment maturity, despite 87% of CFOs calling AI “very or extremely important” to operations in 2026 (Deloitte Q4 2025 CFO Signals, n=200). The gap is execution, not awareness. research/05-analyst-firms/cfo-ai-workflow-automation-finance-function.md
AI reduces the monthly financial close by 7.5 days and enables accountants to serve 55% more clients per week — MIT Sloan and Stanford (Choi & Xie, August 2025, n=277 accountants, 79 SMBs). The only peer-reviewed close-automation dataset in the corpus as of April 2026. research/07-adoption-challenges/ai-cfo-close-process.md
AP month-end workflow compressed from 20 hours to 2 hours at Fanatics Betting & Gaming — cited in Bain Capital Ventures CFO survey (n=50, February 2025). L.E.K. Consulting (n=~100 CFOs, December 2025) documents AP invoice processing dropping from 3 hours to 15 minutes. research/07-adoption-challenges/cfo-ai-decision-framework.md
~56% of CFOs prefer embedded AI over point solutions — L.E.K. Consulting OCFO survey (n=~100 CFOs, 2025). The preference reflects execution risk: CFOs want AI that integrates with existing ERP, FP&A, and close management systems rather than requiring new infrastructure. research/05-analyst-firms/cfo-ai-workflow-automation-finance-function.md
Commentary/report writing leads FP&A AI adoption at 57%; variance analysis (30%) and forecasting (28%) lag — Accounting Today 2025 FP&A survey. The two highest-analytical-value workflows are the least deployed, representing the clearest near-term opportunity for CFO AI programs. research/07-adoption-challenges/ai-cfo-close-process.md
CFO 3-year AI TCO model — framework for modeling total cost of ownership across close automation, FP&A, and AP/AR workflows, with payback benchmarks by company size and ERP environment. research/07-adoption-challenges/cfo-3-year-ai-tco-model.md

KPMG AI in Finance 2026 (May 2026)

KPMG’s May 2026 finance-specific primary survey (n=1,013 senior finance leaders, 20 countries, 13 sectors) establishes the current deployment pace:

93% of US companies plan to deploy or scale AI in finance within 18 months — the shift from “evaluating” to “executing” is now near-universal among companies with budget to act
74% report ROI meeting or exceeding expectations (self-reported; treat as directional, not audited)
50% planning multi-agent AI orchestration across finance workflows — a risk-governance escalation most mid-market finance teams are not yet prepared for
64% cite lack of role-specific use cases and 61% lack hands-on practice environments as the primary adoption barriers
48% worried about accuracy of AI-generated financial outputs — the constraint that matters most in a function where errors carry audit, legal, and investor consequences

Source caveat: KPMG sells AI advisory and finance transformation services; self-commissioned. Large n adds credibility to deployment pace figures; ROI figures require cross-reference with independent evidence.

research/04-consulting-firms/kpmg-ai-in-finance-2026.md

Oliver Wyman Forum / NYSE CFO Agenda 2026 — Finance Workforce Architecture Signal

Oliver Wyman’s inaugural CFO Agenda survey (n=~500 CFOs representing 12% of global market cap, April 2026) provides the most direct evidence yet on how CFOs expect AI to reshape the finance function headcount model:

30% project finance headcount decreasing by more than 10%; 61% expect stability; only 9% forecast meaningful growth
64% anticipate a shift away from junior roles — the clearest single signal that AI is restructuring finance job families rather than simply eliminating headcount
6% rank AI as their primary enterprise value lever despite increasing investment — the accountability problem at the finance chair: CFOs cannot attach P&L outcomes to AI spend when deployment workflows sit in other functions
CFOs at AI-leading firms are nearly 2x more likely to identify as “enterprise insights partners” — strategy and analytics rather than control and reporting — and 80% expect that mandate to grow

The workforce arithmetic: if headcount is stable but 64% shift away from junior roles, mid-to-senior analytical roles must expand. Finance teams are not shrinking; they are upgrading their composition. This requires deliberate training and role-redesign investment, not just a junior hiring freeze.

Source: research/04-consulting-firms/oliver-wyman-cfo-agenda-2026.md · Apr 2026

KPMG Global AI in Finance 2026

KPMG’s global primary survey (n=1,013 senior finance leaders, 20 countries, 13 sectors, revenues ≥$250M, March 2026 fieldwork) is the largest finance-function-specific AI dataset in the 2026 corpus. It establishes current adoption pace and identifies where value accrues:

75%+ of organizations now leverage AI in financial planning, reporting, and commercial analysis — up from roughly 30% two years ago. Finance AI adoption is no longer early-adopter territory; it is the operational baseline for large-company finance functions.
Active AI use more than doubled in two years (from ~30% in 2024) — the pace of adoption has been rapid, but the satisfaction-to-value gap persists: only 23% say AI is exceeding ROI expectations, even though 71% report it is meeting or exceeding them.
Agentic AI deployers outperform peers by 32 percentage points on average (nearly 40pp on forecast accuracy and ROI specifically) — the most significant performance differential in the finance-function AI corpus.
Top gains reported: decision-making speed (71%), decision-making quality (70%), forecasting accuracy (64%) — operational metrics that precede, but do not guarantee, income statement impact. Cross-reference with NBER w34984, which finds a 3x gap between CFO-perceived and revenue-implied productivity gains.
Data fluency named most critical skill need — assessing data quality, interpreting AI outputs, and communicating findings. The finance-function skill gap is not technical proficiency; it is analytical judgment applied to AI-generated outputs.

Source: research/04-consulting-firms/kpmg-global-ai-in-finance-2026.md

IBM IBV “Finance Execution Unlocks AI Value at Scale” (May 2026, n=1,025)

IBM IBV’s global survey of 1,025 finance leaders (published May 26, 2026) — distinct from the earlier IBM IBV + Oracle Dynamic Finance study (n=600, Q4 2025) — establishes a concrete execution-vs-adoption gap:

Median 8% finance cost reduction for experienced AI adopters; that number rises to 18% when AI is embedded end-to-end across the finance function. The 10-percentage-point gap is entirely explained by execution maturity, not model quality.
69% of CFOs call AI integral to finance transformation strategy; >80% rate AI adoption in FP&A and procure-to-pay as important. Conviction is ahead of execution infrastructure.
IBM’s analysis finds that high-performing finance organizations redesign workflows before embedding AI — clarifying decision points, standardizing inputs, reducing handoffs, and defining how AI outputs trigger action. Organizations that deploy AI into existing processes get the 8% result.
The report’s framing (“from experimentation to execution — why sequencing matters”) is consistent with findings across the corpus: McKinsey (only 21% have done workflow redesign), WEF/Accenture (only ~15% have redesigned end-to-end), IBM IBV From AI Projects to Profits (top 18% ROI cohort has completed workflow redesign; median 7% has not).

Source: research/04-consulting-firms/ibm-ibv-finance-execution-ai-scale-2026.md — MEDIUM / TIER 1