The State of Play

AI-driven financial forecasting and scenario modelling is a proven practice stuck between accelerating adoption and constrained execution. Platforms are mature and widely adopted (75%+ of organizations using AI in financial planning per KPMG June 2026), with active deployments at Fortune 500 scale delivering measurable accuracy improvements (64% report improved forecast accuracy per KPMG). Deployment has transitioned from pilots to embedded production: testing/piloting fell from 74% (2024) to <5% (2026) while fully embedded AI rose to 97%, and 76% of finance leaders report seeing ROI within 12 months. Yet this maturation masks an execution-impact gap: only 7% of CFOs report strong business impact despite 60% running AI tools, and successful organizations systematically separate themselves through governance discipline—workflow ownership correlates with 32-point performance advantage (KPMG). Governance and trust remain barriers: practitioner skepticism about output quality, pace of implementation without controls, and mounting concerns about hallucination risk (4.2%-19.1% documented across frontier models) delay forecasting adoption in mainstream organizations. Frontier model reliability constrains autonomous use—GPT-5.5 achieves only 52% accuracy on financial analysis tasks with multi-step reasoning failures. The practice is good-practice because platforms work at scale and CFOs are investing heavily, but tier maturity is constrained by the governance-capability gap, execution discipline requirements, and organizational readiness variation across the market.

Q2–Q3 2026 market signals show agentic deployments at scale yet governance and trust remain the binding constraints on mainstream adoption. Vendor platforms shipped major AI enhancements and reached critical adoption scale: Workday Q1 FY2027 earnings (ended Apr 2026) reports 4,000+ customers now using ≥1 agentic AI product (doubled QoQ) with new ACV from agentic solutions growing 200%+ YoY and approaching $500M ARR; Workday Adaptive Decision Intelligence (GA June 2026) enables natural-language scenario modeling with deterministic calculations, variance analysis, and Monte Carlo simulation for 7,000+ customers; Anaplan released CoModeler and Custom Analyst agents; Board released FP&A Agent. Adoption breadth accelerating: KPMG August 2026 survey (1,013 finance leaders, 20 countries) reports 75% of organizations actively using AI in financial planning (up from 30% in 2024) with 64% citing improved forecast accuracy; agentic deployments show 32-40pp advantage over early-stage adoption; Vena Solutions survey (431 finance leaders) shows 86% actively using AI tools with 34% having fully integrated AI agents across FP&A. Deployment maturity signals advanced: embedded AI deployment rose from 26% (2024) to 97% (2026), testing/piloting declined from 74% to <5%, and 76% report ROI within 12 months. Yet the adoption-impact gap persists acutely: only 7% of CFOs report strong business impact despite 60% running AI; only 12% have forecasting AI in production while 53% don't use AI for forecasting; Gartner's May 2026 survey identifies financial forecasting as among the lowest-rated use cases despite 66% overall efficiency gains. Production deployments include Nasdaq market twins (generative AI stress-testing for limit order books), BARC-verified OneStream user satisfaction (90% rate forecasting highly, 7.0/10 business benefits vs 3.9 for Excel), and workflow accelerations (SaaS consolidation 10 days→3-4 days, professional services budgeting 6 weeks→<3 weeks, Arcis Golf forecasting accuracy +30%). Data architecture remains the binding constraint: FinanceBench analysis documents 79% accuracy with full operational context but 9% without—data grounding, not model capability, determines success. Governance discipline correlates with performance: KPMG research shows workflow ownership and controls separate leaders from laggards by 32 percentage points on forecast accuracy outcomes. Organizational readiness remains uneven: governance concerns (69% of accountants report AI pace without controls), output quality skepticism, and widespread trust erosion slow adoption in mainstream organizations despite vendor advancement and top-line CFO commitment.

Yet frontier model reliability remains a hard constraint on autonomous use, compounded by architectural barriers and governance failures beyond model performance. May 2026 benchmarking by Vals AI shows GPT-5.5 achieves only 52% accuracy on financial analysis workflows, with multi-step numerical reasoning failing below 35% accuracy for sequences exceeding five steps and hallucinated financial figures persisting as a production risk. Peer-reviewed evidence intensifies the signal: July 2026 arxiv research documents 15-23% hallucination rates among high-confidence LLM financial answers, with existing detection methods failing (AUROC 0.55-0.63); independent analysis quantifies impact at $2.3B in trading losses Q1 2026 alone from forecasting AI failures across the industry. Independent testing documents 4.2%-19.1% hallucination rates across frontier models. Critically, hallucination failures now extend across professional services—August 2026 evidence documents major audit/consulting firms (Deloitte Australia, EY) publishing client reports containing fabricated citations, nonexistent studies, and hallucinated references—and subsequent pattern evidence (PwC, KPMG withdrew reports over fabricated achievements and citations) indicates systematic governance failures across Big Four firms, suggesting that even organizations with sophisticated review cultures cannot catch AI errors without process redesign. More fundamentally, Aleph’s analysis identifies the "80% problem"—LLMs are probabilistic (best-guess), but finance requires deterministic outputs (99%+ accuracy, same input always produces same output, defensible to source). This gap demands auditable data layers before tool selection: finance teams succeeding invest in data infrastructure and governance first, then select technology, rather than deploying LLMs against fragmented systems. Empirical research shows the real constraint: financial AI achieves 79% accuracy with full operational context but fails entirely (9% accuracy) with GL data alone, identifying data grounding as the binding lever—not model capability. CFO testing reveals the capability-application gap: Claude and Copilot built 5-year financial models in 15 minutes from single prompts, but with formula errors and structural mistakes requiring expert audit—87% of CFOs expect AI very important yet only 17% actively use it in core workflows (CFO Connect event recap, May 2026). The accuracy problem translates directly to organisational barriers: only 43% of FP&A leaders forecast within 10% accuracy, with 51% ranking accuracy improvement as top-5 2026 priority; AI implementations deliver 15-30% improvements yet most organisations still apply algorithmic forecasts as decision-support only, not primary authority. Deployment reality check: independent field research (August 2026, HCLTech) documents that approximately 95% of enterprise generative AI pilots deliver zero measurable profit-and-loss impact, with deployment friction and organizational adoption (not model capability) identified as the binding constraint. This "Deployment Wall" finding reframes vendor platform maturity as necessary but insufficient—architectural and organizational barriers prevent most pilots from reaching production with realized value. Forrester’s May 2026 predictions underscore the vendor-reality gap: enterprises will defer 25% of planned AI spending to 2027 as the gap between vendor promises and delivered value widens, with fewer than one-third of decision-makers able to tie AI to financial growth. Regulatory overhead layers complexity: NIST AI 600-1 now formally treats confabulation as a tier-1 financial services risk with mandatory pre-deployment testing requirements; by July 2026, FINRA and Treasury released governance frameworks requiring robust pre-deployment accuracy testing and ongoing monitoring for forecasting AI, and federal regulators (Federal Reserve VP) escalated oversight of AI deployment at financial institutions. Critical adoption signal: KPMG Global AI Pulse Q2 2026 shows only 7% of enterprises establish measurable ROI from AI investments; productivity gains declined from 42% to 35% quarter-over-quarter; only 22% can link AI spending to financial outcomes—despite 60% running AI tools. Governance consensus hardens: FMI survey (63 global council members) shows not a single respondent confident in AI forecasting models without human review; 70% use AI in ≤25% of forecasting workflow. Bedford Consulting analysis (August 2026) finds Gartner projects 40% of agentic AI projects will be cancelled by end of 2027, driven by organizational barriers (cost clarity, ROI visibility, control requirements) rather than technical capability gaps—underscoring that deployment readiness, not platform maturity, determines success. The result is market segmentation: well-resourced institutions with enterprise-grade data governance and verified governance frameworks operationalise scenario modelling for competitive advantage (CFOs in board meetings modeling supply chain and headcount changes in real time per FutureCFO May 2026), while mainstream finance organizations stall on hallucination risk, model accuracy uncertainty, regulatory burden, data quality debt, legacy infrastructure constraints, governance escalation, and uneven organizational readiness despite accelerating investment intent and positive ROI reporting among deployed cases.

• 2019: MIT research demonstrates ML superiority over human analysts in earnings forecasting; enterprise planning platforms (Workday, Anaplan) achieve production deployments at Fortune 500 scale; adoption barriers remain structural (human reluctance, control concerns, risk tolerance).
• 2020: Workday Adaptive Planning deployments accelerate across energy, healthcare, finance, and education (ENMAX, Wellcome Sanger, UVA); documented cycle-time reductions of 30–50% and enhanced scenario modelling become standard outcomes. Critical analysis reveals strategic CPM tools lack mature prescriptive analytics for complex manufacturing use cases. COVID-19 drives urgent budget reforecasting demand, accelerating platform adoption.
• 2021: Industry surveys confirm driver-based planning adoption among 342 FP&A professionals and 270 AI models in production at major financial services firms; academic research identifies critical pitfalls in applying ML to causal planning problems. Infrastructure and data governance challenges persist—77% of AI models fail to reach production—constraining scaling of forecasting AI despite platform maturity.
• 2022-H1: Workday reports ~1,500 finance solution deployments including Adaptive Planning with ML-powered forecasting and driver-based scenario modeling; named adopters confirm production adoption at scale. Academic research clarifies forecasting vs. causal planning distinction; survey data reveals adoption gaps—fewer than half of finance professionals prioritize scenario modeling despite technical maturity. Organizational readiness remains the primary bottleneck.
• 2022-H2: Multiple H2 2022 deployments confirm sustained momentum: Sport Alliance rapid Adaptive Planning rollout (Dec 2022 budget release), Rohlik Anaplan with driver-based forecasting (€220M Series D), Swarovski monthly rolling forecasts, retail/distribution 60% planning-time reduction. Forrester TEI validates continued ROI. Critical assessment from Vodafone UK highlights persistent barriers: 24+ months data quality requirement, multi-month transformation effort, cautious organizational approach to production ML. Infrastructure challenges persist—77% of AI initiatives still fail beyond POC.
• 2023-H1: Unilever USA deployment confirms continued production adoption of Anaplan with integrated forecasting processes; rolling forecasts and driver-based planning emerge as standard FP&A landscape expectations. Critical assessment from Babson College researcher identifies finance lagging in AI adoption despite technical maturity. Organizational barriers to algorithmic forecasting adoption persist.
• 2023-H2: Forrester TEI validation confirms 249% ROI from Workday Adaptive Planning deployments. Generative AI enters mainstream finance conversation—IIF-EY survey shows 86% of financial institutions expect significant AI model growth. December CFO survey reveals persistent adoption gap: 42% have not implemented AI despite 83% recognizing importance; 67% of adopters use forecasting. Critical assessment identifies systematic underestimation of climate risks in production scenario models. AI governance and model-risk infrastructure emerge as primary adoption constraints.
• 2024-Q1: Platform deployments continue: Capstone Industries, CU Boulder, and multinational insurance firm adopt Anaplan and Workday Adaptive Planning for budgeting and forecasting. Accenture data shows AI mentions on earnings calls grew from 500 (Q1 2022) to 30,000+ (Q3 2023). SEC enforcement action against Delphia and Global Predictions for false AI claims exposes governance risks in algorithmic forecasting. Practitioner assessments document generative AI limitations in scenario edge-case thinking despite mainstream interest in GenAI forecasting capabilities.
• 2024-Q2: CU Boulder's Anaplan deployment goes live with budgeting and compensation planning; Gartner survey shows 66% of finance leaders expect GenAI to massively impact forecast/variance explanations. Oracle support documentation reveals persistent accuracy and interpretability issues in production forecasting systems. Expectations for GenAI in forecasting accelerate, yet real-world deployments continue to face data quality, explainability, and organizational adoption barriers. Vendor platform maturity confirmed; algorithmic forecasting remains decision-support rather than primary authority.
• 2024-Q3: Gartner survey confirms 58% adoption of AI in finance functions, with 28% using analytics for forecasting—mainstream adoption signal. Yet FP&A practitioner surveys reveal adoption-execution gap: 78% of teams struggle to run scenarios within a day; 70% still rely on spreadsheets; only 9% use driver-based models. VEIC (sustainable energy) deploys Workday Adaptive Planning for monthly forecasting and scenario modeling in September. Critical assessments emerge: MIT economist Acemoglu questions AI productivity claims; Gartner projects 30% of GenAI projects will be abandoned by 2026 due to data quality and ROI challenges. TechCrunch highlights paradox of ROI measurement in AI financial operations. Consensus view: tools mature and deployed at scale, but organizational adoption gaps and data quality constraints remain primary bottlenecks.
• 2024-Q4: Divergence between vendor claims and practitioner reality sharpens. KPMG survey reports 78% of US finance leaders piloting/using AI for planning with 92% meeting ROI expectations, yet FP&A Trends survey of 2,400+ practitioners shows only 6% AI adoption, 22% able to run scenarios same-day, 52% still using Excel. Anaplan releases PlanIQ general availability for predictive forecasting with AWS integration. Birch Family Services (NYC non-profit) deploys Workday Adaptive Planning, reducing budgeting from 3-5 months to 2 months. Critical assessments intensify: research shows GPT-4 earnings forecasts less accurate than human analysts; Economist Impact survey finds 85% enterprises testing GenAI but only 22% confident in IT architecture and 60% UK firms with zero production GenAI deployment. Architecture, data governance, and production readiness emerge as primary barriers to realizing vendor platform capabilities.
• 2025-Q1: Platform vendors advance agentic AI capabilities: Anaplan launches CoModeler and Finance Analyst agents for scenario modeling and planning. Research identifies technical solution—GenAI synthetic data addresses ML overfitting in forecasts and enables robust scenario exploration. However, practitioner adoption remains constrained: FP&A Trends 2025 survey shows only 6% AI adoption despite 23pp quality improvement for users; only 5% of companies use AI for financial decisions. Investment momentum continues—all 56 financial institutions increase AI/ML spend—but deployment risks surface: LLM hallucinations in 41% of financial queries raise confidence barriers, while only 5% of companies operationalize AI-driven forecasting. Organizational readiness and trust in algorithmic outputs remain primary constraints despite platform maturity.
• 2025-Q2: Anaplan Intelligence white paper details expanded agentic capabilities; House of HR and other mid-market firms deploy Anaplan for FP&A and scenario planning. Evidence of realized value emerges: global manufacturer achieves 22% capital efficiency gain via AI scenario modeling. However, independent BARC analyst review reveals adoption-execution gap: 2,400+ Anaplan customers show strong user experience (8.9/10) but mixed satisfaction (5.4/10) and modest business value perception (7.4/10). Critical assessments intensify: Guidehouse analysis documents only 4% achieve significant AI returns, with 30% of GenAI projects abandoned post-POC; financial services practitioners cite regulatory scrutiny, control requirements, and data governance as primary deployment barriers. Landscape splits between early-adopter successes and mainstream market paralysis.
• 2025-Q3: Platform adoption accelerates: Deepak Fertilizers' five-year Anaplan deployment demonstrates manufacturing-sector integration; Protiviti research shows AI adoption in finance more than doubled YoY with scenario planning as CFO priority. Regulatory drivers emerge: OSFI-AMF climate scenario exercise with 250+ Canadian financial institutions and Moody's analysis of quantitative risk appetite frameworks signal institutional mandate-driven scenario modeling adoption. Critical tension surfaces: Workday Adaptive Planning shows 7,000+ customers globally with embedded AI/ML, yet independent BARC assessment reveals moderate satisfaction (5.5/10); 85% of companies miss AI-driven cost forecasts by >10%, exposing accuracy and visibility gaps in production deployments. Market segmentation sharpens: large financial institutions integrate regulatory scenario modeling effectively, while mid-market remains challenged by data governance and organizational readiness.
• 2025-Q4: CFO sentiment shifts toward transformation maturity and governance rigor. Microsoft/IDC analysis identifies "Frontier Firms" (financial services orgs embedding AI agents across workflows) reporting 3x higher ROI than slow adopters, with scenario modeling as differentiator for strategic finance. FSB regulatory monitoring confirms widespread institutional AI adoption in financial sector but flags third-party dependencies and governance vulnerabilities. Paystand survey shows FP&A remains top AI disruption area (44%), though 65% of finance teams still in exploratory phase. Platform user satisfaction remains high (84-93% likeliness to recommend for Anaplan/Workday), yet Pertama Partners' December analysis documents persistent ROI reality gap: 68% of AI projects fail to meet ROI expectations with actual returns 47% below projections—integration costs underestimated, adoption overestimated. Year-end CFO sentiment (Fortune interviews) emphasizes validation, governance, clean data, and enterprise-grade architecture as 2026 priorities. By end of Q4, the landscape clarifies: platforms achieve genuine enterprise adoption scale and regulatory legitimacy; CFO commitment to AI transformation intensifies; yet practitioner-level operational barriers (cost visibility, forecasting accuracy, change management) persist, creating widening gap between top-down transformation mandate and bottom-up execution capability.
• 2026-Jan: Early 2026 signals reveal persistent execution barriers despite accelerating adoption intent. Mid-market CFO survey (100 organizations) shows 60-77% plan to adopt AI for finance, yet only 14% trust AI for accuracy without human oversight—demand for explainability and "intelligent escalation" (autonomous on routine, human oversight on exceptions) hardens. Investment prioritization strengthens: 64% of finance leaders rank AI/ML as leading technology priority. However, critical assessments document systemic challenges: MIT research reveals 95% failure rate for enterprise GenAI projects with no measurable P&L impact within 6 months; 77% of enterprises cannot measure ROI despite AI deployment. Skills gaps persist as primary organizational barrier: CIMA survey shows 88% of finance leaders expect AI to transform profession within 1-2 years but 50% cite skills deficit and 41% cite organizational coordination challenges. Practitioner insights from scenario planning consultancy confirm AI useful for driver ideation and narrative creation but outputs remain bland without human supervision. Critical analysis of financial forecasting acknowledges AI potential for scale and NLP sophistication but highlights risks: false precision in point estimates, correlated models creating echo chambers, narrative feedback loops where companies adapt disclosure language to AI-driven expectations. Consensus view entering 2026: adoption acceleration visible but execution remains the constraint—vendor platforms mature and widely deployed, CFO transformation commitment firm, yet operationalization barriers (cost visibility, forecasting accuracy, trust) and widespread project failures keep practical AI-driven forecasting limited to Frontier Firms with enterprise-grade architecture and data governance.
• 2026-Feb: Platform adoption breadth accelerates while value realization concentrates among well-resourced organizations. Finastra survey documents 65% of US financial institutions in active AI deployment with 42% planning >50% investment increase; Broadridge study shows 80% of financial services firms using generative or predictive AI. However, critical reality-gap widens: only 27% report measurable business benefits; MIT analysis confirms only 5% of enterprise AI pilots deliver measurable value; Deloitte data shows only 6% achieve ROI within a year. FP&A-specific barriers sharpen: Gartner projects >40% of agentic AI projects cancelled by 2027; AI modeling failures emerge (hallucinations, circular logic collapse) requiring hybrid analyst validation; 37% of reported time savings consumed by fixing AI outputs. Schellman case study demonstrates production-stage Workday Adaptive Planning deployment with expanded capabilities (workforce planning, reporting, revenue modeling). Consensus clarifies: adoption intent and investment commitment intensify, yet practical operationalization remains confined to organizations with enterprise-grade data governance and third-party oversight—execution barriers override platform maturity.
• 2026-Q1: Vendor innovation accelerates but value-delivery gap persists. Workday Adaptive Planning 2026 R1 (March) expands Predictive Forecaster to 10M cells, launches Planning Hubs for consolidated workflows, adds ML-driven anomaly detection. Anaplan launches CoModeler, Custom Analyst, and Agent Studio agents combining LLMs with deterministic planning engine for auditable scenario modeling (March). Board releases FP&A Agent with econometric forecasting claiming 50% forecast accuracy improvement via 5M+ economic signals. Virtasant case analysis: median AI ROI in finance at 10% despite 72% adoption; Coca-Cola reduced treasury workload 14% via AI forecasting. JPMorgan case: 450+ GenAI use cases in production including treasury stress scenarios and scenario analysis with firmwide CDO governance. Yet implementation barriers remain acute: Fullstack Labs analysis of 140 GenAI implementations shows 73% ROI failure; mid-market CFO survey shows only 14% trust AI for accuracy without human oversight; practitioner data shows 78% of teams cannot run scenarios same-day, 64% find scenario planning extremely challenging. Critical assessment documents recurring Adaptive Planning implementation failures—model configuration breaking, data governance insufficient, structural complexity management problems—signaling persistent operational barriers despite platform maturity and accelerating investment.
• 2026-Apr: Execution failures dominate the signal: Gartner finds two-thirds of finance AI buyers experience post-purchase regret, and FP&A Trends confirms only 3% of organizations achieve real-time scenario capability while 64% cite scenario planning as their most difficult process. Technical analysis clarifies the architecture divide — traditional ML achieves 95-98% accuracy on bookings forecasts while LLMs fail due to hallucinations and context limits, with VeNRA's neuro-symbolic approach (1.2% hallucination rate) emerging as a reliability candidate. PwC data (1,217 executives) shows 74% of AI value is captured by just 20% of organizations, with 56% reporting zero financial benefit — a concentration dynamic directly shaping who benefits from Workday Adaptive Planning's 2026 R1 (10x predictive forecasting scale, planning hubs) and Anaplan's named-customer validation (Sky, Virgin Media O2). Duke/Federal Reserve peer-reviewed research confirms the productivity paradox: CFO-reported gains of 1.8% exceed what revenue outcomes imply, indicating an execution lag that keeps the practice stalled despite Oracle embedding Advanced Predictions ML into EPM at no additional cost and Anaplan pricing pressure driving mid-market evaluation of faster-deploying alternatives.
• 2026-May: Vendor ROI evidence intensified alongside sharpening technical constraints and an adoption inflection signal. Consero Global survey (102 PE/VC-backed CFOs) shows 42% AI deployment (up from 22% YoY) with 65% closing in under 9 days — up 8x from 2024 — while 43% of FP&A leaders still miss 10% accuracy targets and 51% rank accuracy improvement as their top-five 2026 priority. Workday Adaptive Planning Forrester TEI documented 242% ROI, $6.3M 3-year benefits, and 35% FP&A productivity gains; a NYSE/Oliver Wyman survey of 500 CFOs (12% of global market cap) confirmed only 8% have deployed AI agents at scale while 74% remain in planning or piloting. Hallucination risk hardened as a structural barrier: ACL 2026 peer-reviewed research (FinGround) proved existing detectors miss 43% of computational errors and achieved only 68-78% hallucination reduction; Vals AI Finance Agent v2 benchmark showed GPT-5.5 at 52% accuracy on financial analysis tasks with multi-step numerical reasoning failing below 35% accuracy for sequences exceeding five steps; an independent 5-model benchmark documented 4.2-19.1% hallucination rates with citation accuracy worst at 12.4%; and NIST AI 600-1 formally classified confabulation as a tier-1 financial services risk requiring mandatory pre-deployment testing — adding governance overhead to already-complex deployments. Forrester's May 2026 predictions confirmed enterprises will defer 25% of planned AI spending to 2027 as the vendor-promise gap widens, with fewer than one-third of decision-makers able to tie AI to financial growth.
• 2026-Jun: Cross-vendor and survey signals converge on a persistent data-architecture constraint as the binding limit for AI forecasting accuracy, alongside striking deployment maturation metrics. KPMG 2026 (1,013 senior finance leaders, 20 countries) confirms 75%+ leveraging AI in financial planning with agentic deployments showing a 32-40 percentage point advantage on forecast accuracy — yet Gartner's concurrent survey of 204 finance leaders identifies financial forecasting as among the lowest-rated use cases despite 66% reporting efficiency gains, underscoring the capability-value gap. Vena Solutions survey (431 finance leaders) shows 86% actively using AI tools with 34% having fully integrated AI agents across FP&A; a separate CFO benchmarking study documents that enterprise AI embedding jumped from 26% to 97% between 2024 and 2026 while testing/piloting fell from 74% to <5% — deployment transition from pilot to production is now essentially complete at scale. Workday shipped Adaptive Decision Intelligence (natural-language scenario modeling, GA) for non-technical decision-makers, while ChatFin's production analysis establishes that AI achieves 2% daily forecasting error with clean operational data but fails entirely with GL data alone — framing data architecture, not model capability, as the binding constraint. ACCA/IMA Q1 2026 survey of practitioners documents three persistent governance headwinds — implementation pace without controls, output quality concerns, and governance erosion — indicating that the trust gap constraining mainstream adoption has not closed despite vendor maturation. Galorath's industry survey (220 professionals) reports 79.1% increased AI spend on estimation with 51% of aggressive adopters reporting significant accuracy improvements, confirming continued investment momentum despite execution barriers.
• 2026-Jul: Platform-level integrations and reforecasting urgency define the period: Pigment's MCP integration with Claude reduced multi-week scenario-modeling workflows to one week in early deployments, while Workday's 200% AI agent ARR growth (4,000+ active customers, 50% of transactions involving AI) signals that production-scale FP&A automation is now commercially mainstream at category leaders. Governance barriers remain the binding limit: Gartner identifies financial forecasting as among the lowest-rated AI use cases despite 75% of finance leaders using AI, and FINRA/Treasury released governance frameworks requiring robust pre-deployment accuracy testing and ongoing monitoring for forecasting AI — while Richmond Fed CFO survey data (474+ respondents) shows reforecasting activity jumped 63% quarterly, confirming urgent demand that most organizations lack the infrastructure to satisfy reliably. A University of York peer-reviewed study of 21,000+ firms over 55 years found ML models achieve 70% accuracy on earnings prediction with portfolios outperforming human-analyst benchmarks by 7-10% annually, even as hallucination costs mounted — independent analysis quantified $2.3B in Q1 2026 trading losses attributable to forecasting AI failures, with 82% of production AI bugs traced to hallucinations. Workday shipped native MCP server integration for Claude/ChatGPT within Adaptive Planning (joining Pigment's earlier MCP deployment), while FinanceBench research confirmed the data-grounding constraint driving these failures: 79% accuracy with full operational context collapses to 9% without it. Governance-accountability gaps sharpened further: Anrok analysis found 66% of CFOs require human oversight for agentic AI with over 80% encountering hallucinations, Deloitte's Q2 2026 CFO Signals (200 US CFOs) recorded only 43% governance confidence despite 93% organisational AI use, and Avalara's 1,505-CFO survey confirmed 92% face pressure to justify AI ROI with 44% unable to explain AI actions to auditors. Cambridge Judge Business School's maturity model quantified the adoption-transformation gap (81% adopting AI, only 40% at advanced stage, 68% citing legacy systems as the primary 12-18 month bottleneck), even as new named deployments kept validating platform value — a QMetrix review of 15+ Workday Adaptive Planning customers documented rolling forecasts 75% faster and P&Ls generated in minutes versus 48 hours, and an Asia-based financial services firm's stochastic scenario-modelling pilot outperformed analyst forecasts by 22%.
• 2026-Aug: KPMG's 20-country survey confirmed AI use in finance doubled to 75% (from 30% in 2024) with 64% reporting improved forecasting accuracy, while named deployments (Arcis Golf's 30% forecasting-accuracy gain via Workday, Adaptive Planning consolidation cases cutting forecast cycles from days to hours) and Workday's doubling agentic-AI customer base (4,000+, ACV growth 200%+ YoY) signal continued vendor momentum. Countervailing evidence hardened: HCLTech field research found ~95% of generative AI pilots deliver zero P&L impact, Gartner projects 40% of agentic planning projects will be cancelled by 2027, and further Big Four hallucination cases (Deloitte, EY, KPMG, PwC) underscored that organizational readiness—not platform capability—remains the binding constraint.