← Knowledge Base 🕐 37 min read
Knowledge Base

AI Budget and CFO Decision-Making

· March 2026

The structured process by which a mid-market CFO allocates, tracks, and defends AI spending across a multi-year arc. ...

The structured process by which a mid-market CFO allocates, tracks, and defends AI spending across a multi-year arc. As of April 2026, AI represents 41.5% of all new technology spending globally (Gartner, Feb 2026). The average organization’s AI spend hit $85,521 per month in 2025 — a 36% year-over-year increase — yet only 51% can evaluate whether those dollars deliver returns (CloudZero State of AI Costs, n=500 U.S. software leaders, March 2025). The CFO’s problem is no longer whether to fund AI; it is how to fund it without budgeting for one year while committing to three.

Everyday AI Ep 755 — The 88%/6% Adoption-vs-Profit Gap (April 2026)

Everyday AI Ep 755 “Managing the AI Capability Gap” (Jordan Wilson, April 14, 2026) cites McKinsey data that provides the starkest single-sentence framing of the AI budget accountability problem:

  • 88% of organizations claim regular AI use; only 6% generate meaningful profit from AI (McKinsey, 2026). For CFOs tracking AI program investment: if the organization is in the 88% that “uses AI” but not the 6% generating profit, the budget question is not “should we invest more?” but “are we investing in the right things?”
  • Anthropic labor data adds structural context: AI could theoretically automate 94% of computing and math tasks, yet observed usage sits at 33%. In management, finance, and legal roles — the CFO’s primary domain — theoretical capability is 80–90% but actual application is under 20%. The deployment gap is not a model quality problem; it is a workflow redesign problem.
  • Leading organizations allocate 20% of digital budgets to rebuild knowledge workflows rather than to purchase additional tools. The CFO who frames AI investment as tool licensing is systematically misallocating relative to the cohort generating returns.
  • The primary metric for tracking whether an AI program is in the 6% versus the 82% residual: percentage of AI-assisted workflow steps accepted without rework, segmented by risk level — a measurable operational indicator that precedes financial returns by 6–12 months.

Source: research/13-multimodal-sources/everyday-ai/2026-05-22-ep755-780-enterprise-ai-mining.md · Everyday AI podcast, Jordan Wilson · Ep 755 (Apr 14, 2026) · McKinsey data cited at HIGH source credibility; episode interpretation MEDIUM · TIER 1

Deloitte “AI Tokenomics” + Benchmarkit: The Forecasting Gap (Apr 2026, n=550 + n=372)

  • 80% of enterprises miss AI infrastructure forecasts by more than 25% — only 15% forecast within a 10% margin (Benchmarkit/Mavvrik, n=372, Sep 2025); errors are nearly always in one direction: too low
  • The token cost paradox: per-token costs fell 280x from 2023 to 2026, yet average enterprise AI budgets grew 483% ($1.2M→$7M). Volume overwhelms unit economics — usage growth outpaces efficiency gains
  • 84% of organizations report AI infrastructure eroding gross margins 6%+; 26% see erosion exceeding 16% — margins are the canary, not a projection
  • LLM tokens rank fifth among AI surprise costs. Data platforms (56% cite as #1 surprise), network access (52%), integration labor, and infrastructure over-provisioning all hit harder and earlier than the vendor API bill
  • Agentic AI is the unmodeled multiplier: chatbot = 1x tokens; RAG query = 3–5x; agentic workflow = 10–20x per task. Organizations that approved AI assistant budgets and then expanded to agents absorbed the multiplier without reforecasting
  • Chargeback = 2x cost maturity: organizations that charge business units for AI consumption demonstrate 2x greater cost management maturity than those treating AI as shared infrastructure

Source: research/05-analyst-firms/deloitte-ai-tokenomics-cfo-guide-2026.md · Apr 2026

Why this matters to mid-market buyers

  • License fees are 10–17% of true AI spend. The remaining 83–90% sits in integration, data governance, training, change management, security, maintenance, and consumption-based surcharges that 78% of IT leaders report as unexpected (Zylo 2026 SaaS Management Index; CloudZero, 2025). A budget that only covers licensing is a budget that misses most of the cost.
  • The 3-year total cost for a 500-person mid-market company deploying AI across 3–5 workflows: $550K–$1.4M. That range collapses to $450K–$800K for organizations that invest in data governance and change management in Year Zero, and balloons past $2M for those that skip foundational work.
  • PwC’s 29th Global CEO Survey (n=4,454, January 2026) finds 56% of CEOs report zero revenue or cost improvement from AI. The 12% seeing both lower costs and higher revenue share a common trait: they budgeted for the full three-year cost arc — assessment through scaling — before writing the first check.
  • Deloitte’s 2026 Global Human Capital Trends (Mar 4, 2026, n=9,000+ leaders, 89 countries, Oxford Economics fieldwork) finds that 59% of organizations are running tech-first AI strategies, and that cohort is 1.6x more likely to miss ROI expectations than the human-centric cohort. The companion Tech Trends research reports 93% of tech funding goes to technology itself and only 7% to training and upskilling. A CFO funding a Copilot rollout without funding the workflow redesign that makes Copilot useful is underwriting the 1.6x shortfall with the 93/7 split.

Practitioner voices (pillar 13)

“We went from a target of a million dollars in revenue per employee to a target of about 10 million per employee with AI.”

— Chris Happ, CEO, Virtuous AI · April 2026 · research/13-multimodal-sources/ai-for-the-c-suite/2026-04-14-chris-happ-why-98-of-ceos-know-ai-matters-but-only-7-have-a-.md

“On average in the mid-market, studies will tell you that you have 50 different applications running. So 50 silos of data costing roughly $2,000 per person per year.”

— Chris Happ, CEO, Virtuous AI · April 2026 · research/13-multimodal-sources/ai-for-the-c-suite/2026-04-14-chris-happ-why-98-of-ceos-know-ai-matters-but-only-7-have-a-.md

“Every CIO has some impetus to want to embrace this technology and really put it to work in the enterprise. But fewer than half of them see quick ROI or see ROI on the horizon or see ROI that’s in a very achievable, certain way.”

— Linda Yao, VP of AI Solutions and Services / COO, Lenovo · April 2025 · research/13-multimodal-sources/me-myself-and-ai/2025-04-01-speed-ease-and-expertise-with-ai-lenovos-linda-yao.md

“The budgets for this are not going to magically appear. So you got to manage your costs and what you’re doing and make sure that you find out where you have waste and redeploy it towards some of these solutions.”

— Vlad Lukic, Global Leader for Tech and Digital Advantage, Boston Consulting Group · September 2025 · research/13-multimodal-sources/beyond-the-pilot/2025-09-13-venturebeat-in-conversation-gen-ai-what-the-enterprise-is-ge.md

Lukic’s framing from BCG’s global tech practice matches the cost-reallocation logic in the BCG “AI-First Cost Advantage” section below: AI budgets are not new money, they are reallocated money from waste elimination. A CFO who budgets for AI as a separate line item — rather than funding it from demonstrable cost-saves in the first six months — is creating financial exposure rather than managing it.

BCG “How Leaders Build an AI-First Cost Advantage” (Berthion/Brunelli/Catchlove/Goydan, Mar 26, 2026) — AI and cost transformation as a single, self-funding strategy

  • The framing shift the piece lands is the one that matters most for mid-market CFOs: AI and cost transformation are not two workstreams — they are one sequenced play. Use traditional cost levers (supplier reviews, spec reviews, inventory optimization, marketing analytics, customer service triage) to generate 5–25% near-term savings in 3–6 months, and reinvest those savings to fund the deeper workflow redesign that delivers the 3–4x breakthrough multiplier.
  • The leader-vs-peer differential BCG publishes: 3x greater cost reduction, 1.6x higher EBIT margins, 2.7x the return on invested capital. BCG-defined cohort — apply vendor caveat — but consistent in direction with the McKinsey AI Transformation Manifesto (20% EBITDA uplift, $3 per $1 invested on n=20 leading companies) and the broader 2026 leaders-vs-peers literature.
  • The 10/20/70 CFO lens — 10% of value from algorithms, 20% from technology and data, 70% from process redesign — is the simplest business-case screen a CFO can apply. If an AI proposal puts more than 30% of spend into tooling and less than 60% into workflow redesign, training, change management, and process ownership, it belongs in the 60%-of-companies-report-minimal-value cohort. Send it back.
  • BCG’s four-move roadmap is a budget-sequencing framework: (1) start with proven deployments to fund the journey; (2) reinvent workflows end-to-end for 3–4x the incremental impact; (3) apply agentic AI in low-compliance, high-complexity process environments; (4) rigorously track every efficiency gain to a specific P&L line item with a decision made before rollout whether the gain becomes headcount reduction, capacity reallocation, or employee-morale reinvestment.
  • The most concrete benchmark for a mid-market CFO is IBM’s own cost transformation — >$4.5B in annual operating cost reduction, 90%+ of HR inquiries resolved through AI chatbot with HR opex down 40% and customer loyalty up 74 points, FP&A costs down 35%, IT costs down roughly $600M. That is a vendor case (IBM is BCG’s case subject) — treat the absolute dollars as directional — but the composition (sequence HR and finance automation, reinvest into IT platform consolidation, then attack third-party spend) is the template a 400-person company can scale down.

Source: research/04-consulting-firms/bcg-ai-first-cost-advantage-2026.md

a16z Third Annual CIO Survey (n=100 Global 2000 CIOs, Feb 2026) — The LLM Spend Escalator

  • Average enterprise LLM spend: $4.5M (2024) → $7M (early 2026) → $11.6M projected by year-end — a 158% two-year increase on the model-access line alone, before application development, integration, training, and governance costs. For a $2B revenue company spending at the survey average, LLM spend approaches 0.6% of revenue by year-end 2026.
  • Application spend surprised to the upside: enterprises budgeted ~$3.9M for AI applications and tracked ~$6M actual — a 54% overage driven by consumption-based pricing that is structurally difficult to forecast at scale.
  • 65% of enterprises prefer incumbent solutions (Microsoft, Salesforce, ServiceNow) for integration ease and procurement simplicity — yet this cohort consistently reports lower ROI than organizations running frontier models on purpose-built workflows. The CFO who funds only what procurement recommends is underwriting lower returns.
  • 81% of enterprises now run three or more model families (up from 68% nine months earlier) — multi-model is the default, not a leading-edge choice, and multi-vendor governance adds cost complexity that single-vendor budgets don’t model.

Source: research/01-ai-native-landscape/a16z-enterprise-cio-third-annual-2026.md · Feb 2026 · Note: a16z is a VC firm with portfolio exposure to AI vendors — treat market share data as directional; use Ramp AI Index for calibration.

Oliver Wyman Forum / NYSE CFO Agenda 2026 (n=~500 CFOs, Apr 2026) — the 6% problem

  • Only 6% of CFOs rank AI deployment as their primary lever for driving enterprise value — below growth (64%), cost management (60%), enterprise value creation (65%), and strategic fit (62%). Most CFOs are increasing AI investment while simultaneously treating it as a low-priority value driver: the classic accountability gap from the finance chair.
  • 30% of CFOs project their finance workforce will shrink by more than 10%. 64% expect a shift away from junior roles. Yet 61% expect stable total headcount — the only arithmetic that works requires mid-to-senior analytical roles to expand. Finance functions are not getting smaller; they are getting structurally different.
  • CFOs at AI-leading organizations are nearly twice as likely to identify as “enterprise insights partners” rather than traditional stewards, and 80% expect their strategy and transformation mandate to grow over three years. The CFO role is bifurcating.
  • The accountability gap mechanism: CFOs approve AI budgets without owning deployment workflows (owned by CIO, business units, operations). When AI returns accrue to operations or engineering, the CFO sees the cost side and the benefit side indirectly. The 6% prioritization reflects this structural attribution problem, not CFO disinterest.
  • Companion CEO data (n=415): 27% of CEOs say AI ROI met/exceeded expectations (down from 38%); 53% say too early to assess; 24% report zero revenue impact. The CFO’s 6% prioritization and the CEO’s 27% satisfaction rate are two points on the same curve.

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

McKinsey Global Tech Agenda 2026 (Reil-Jerenz et al., Feb 2026) — the 2026 tech-budget divergence

  • McKinsey’s n=632 C-level primary survey (Sep 29 – Nov 10, 2025, 69 nations / 24 industries) delivers the cleanest quantitative expression of the 2026 performance divide yet published: 28% of top-performing companies plan to increase tech budgets by more than 10% in 2026, versus just 3% of others. Half of all respondents plan increases of at least 4%. AI has surpassed cybersecurity and infrastructure modernization as the top tech transformation priority for the next two years — 50% of all companies, 54% of top performers — name AI a priority investment.
  • The CFO implication is stark. The ≤10% budget bracket is where 66% of non-top-performers sit. The >10% bracket is where 28% of top performers are funding. If a 2026 tech budget is tracking <10% growth, the CFO is not underfunded by a little relative to the cohort delivering ≥10% revenue and EBIT growth — it is an order-of-magnitude gap. That is the benchmark conversation the board will have in Q3 2026 whether the CFO has staged it or not.
  • McKinsey’s top-performer cohort pairs the spend increase with three specific allocations: (1) agentic AI deployment with full operating-model transformation (Aviva’s 80-AI-model claims overhaul: liability assessment −23 days, complaints −65%, CSAT 7x); (2) insourcing/reskilling over outsourcing — half of top performers plan to increase insourcing vs. 37% of others; (3) technology-executive hiring at nearly twice the rate (37% vs. 19%) plus more financial managers to tie tech investment to measurable ROI. The budget increase pays for all three.
  • The CIO-strategy-involvement gradient is the secondary CFO signal. At <$50M in 2025 tech spend, only 48% of CIOs are “very involved” in shaping enterprise strategy; at ≥$500M, 74% are. Scale pulls CIOs into strategy faster than performance does. A CFO sponsoring the tech budget has to decide whether to fund the CIO’s strategic involvement or accept the ceiling the budget bracket implies.
  • Vendor caveat: McKinsey/QuantumBlack has direct commercial interest in CIO-transformation, product-and-platform operating-model, and agentic-AI engagements. The “top performer” cohort is McKinsey-defined (≥10% average growth in both revenue and EBIT over three years, self-reported). Triangulates with BCG AI Radar 2026 (Trailblazers commit 73% of transformation budget to AI vs. 24% for Followers) and IBM IBV Tech Debt Reckoning (+29% ROI uplift when tech debt is priced into AI business cases) — the pattern of top-performer budget concentration is consistent across three institutional surveys with different methodologies.

Source: research/04-consulting-firms/mckinsey-global-tech-agenda-2026.md

Forrester “2026 Really Is This Risky: Our Top Recommendations For CISOs” (Burn + Pollard, Mar 4, 2026) — the AI-security budget-restructure argument

  • Burn and Pollard name the single highest-leverage 2026 security-program move as a budget-restructure: shift AI security costs out of the security budget and into enterprise AI investments. The reasoning is that AI security is not a niche control set; it is a business risk that scales with AI adoption across marketing, operations, and product. Funding it solely from the security budget “guarantees tradeoffs that weaken core defenses.”
  • The CFO implication is concrete. Every new AI business case (agentic customer service, marketing personalization, forecasting automation) should include an “AI controls” line item inside the project budget — covering DLP, non-human identity management, prompt-injection monitoring, and AI-specific tabletop exercises. The CISO no longer has to decide between funding MFA hardware rotation and funding the new agent-monitoring tooling, because those budgets sit in different places and compete with different things.
  • Pairs with the BCG 10/20/70 budget rule already in this hub: where 10/20/70 prescribes how much of the AI budget goes to algorithms vs. technology vs. people/process, Forrester prescribes where the AI-security portion lives inside the finance structure — inside the 20% AI-technology line of the initiative’s own budget, not inside the separate CISO cost center.
  • Direct 2026-budget-cycle exposure: if the 2026 security budget funds AI security as a line item inside the security cost center, the CFO has already authorized cannibalization of foundational defenses the first time the business funds a new AI initiative mid-year. The fix is a Q2 CFO conversation ahead of the next AI business case, not a security-architecture change.

Gartner “Autonomous Business and AI Layoffs Do Not Deliver Returns” (n=350, Q3 2025, May 2026)

  • 80% of organizations that piloted or deployed autonomous AI or intelligent automation reported workforce reductions. Yet workforce reduction rates were nearly identical for high-ROI and low-ROI organizations — layoffs do not predict returns.
  • The CFO business-case implication: headcount reduction is a one-time savings event. It creates budget room; it does not create compounding return. An AI ROI model that rests primarily on a headcount-reduction line is a model the Gartner data does not support.
  • High-ROI organizations invested more aggressively in three areas post-deployment: skills (retraining workers to direct autonomous systems), roles (creating new positions — workflow architects, AI oversight leads — that did not exist before), and operating models (redesigned approval chains and quality controls).
  • Helen Poitevin, Distinguished VP Analyst at Gartner: “Chasing value only through headcount reduction is likely to lead most organizations down a path of limited returns.”
  • Gartner projects autonomous AI will be a net-positive job creator by 2028–2029 as new categories of work emerge.
  • Mid-market implication: sequence the investment in human capability before deploying autonomous systems at scale. The 80% who cut staff and the minority who see genuine ROI are not the same organizations.

Source: research/05-analyst-firms/gartner-autonomous-business-layoffs-roi-2026.md

Source: research/06-security-frontier/forrester-ciso-2026-recommendations.md

McKinsey/Serviceware “Recalibrating CIO Technology Budgets for the AI Era” (~Apr 2026, n=17)

  • AI eats up to one-third of change budgets while simultaneously adding to run costs. The structural squeeze is bidirectional: every AI deployment creates new run obligations unless an equivalent legacy workload is retired.
  • Four IT archetypes: deliberate modernizer (target), strained transformer, lean operator, heavy IT sustainer. Most mid-market companies sit in heavy IT sustainer territory without realizing it.
  • Deliberate modernizer benchmarks: run costs ≥20% below peers; 57% of application spending on modernization and new capability (change); at least one-third of total tech budget allocated to change; 16% of total tech budget on internal staff working on change (1.5–4.0× laggards).
  • CFO implication: a $10M technology budget at the deliberate modernizer target implies $1.6M in internal change capacity, $3.3M+ in change allocation, and a standing legacy-retirement process. Most mid-market CIOs cannot identify their run/change split precisely — that gap is the first diagnostic.
  • Prerequisite lever: explicitly retire legacy applications before approving AI workloads. Without a standing retirement process, every new AI tool compounds run costs rather than replacing them.

Source: research/04-consulting-firms/mckinsey-cio-tech-budget-ai-era-2026.md

EY US AI Pulse Survey Wave 4 (n=500 SVP+, Oct 2025) — the IT budget reallocation signal

  • 27% of organizations already allocate 25%+ of their IT budget to AI; that figure is expected to nearly double to 52% within 12 months. AI is transitioning from a project line to a core infrastructure cost category — CFOs who treat it as a variable line will find it behaving as a fixed one.
  • The ROI threshold gap is quantified: organizations investing $10M+ report significant productivity gains at 71% vs. 52% for those investing less than $10M. The implication for mid-market CFOs: partial investment is not a neutral hedge. It incurs disruption costs without capturing the return.
  • 96% report some productivity gains; only 57% call them significant. The 39% in between are in the most dangerous CFO position: incremental cost exposure without the financial performance that justifies continued investment.
  • Only 17% used productivity gains to reduce headcount. The other 83% reinvested into more AI capability, cybersecurity, R&D, or upskilling — meaning AI productivity savings are not accruing to the P&L on a 12-month cycle. Budget models that project cost savings flowing to the bottom line within the year are structurally misaligned with what senior leaders are actually doing.

Source: research/04-consulting-firms/ey-ai-pulse-survey-wave4-2025.md · MEDIUM / TIER 2 · n=500 US SVP+, 50 per industry, Oct 2025 · EY has commercial interest in AI adoption services.

KPMG “AI in Finance: The Decision Advantage” (n=1,013, May 2026) — finance function deployment acceleration

  • 93% of US companies plan to deploy or scale AI in their finance function within 18 months. The transition from pilot to production is no longer a future planning item — it is the current execution agenda.
  • 74% of finance leaders report AI ROI meeting or exceeding expectations (self-reported against self-set targets; treat as directional confidence signal, not audited performance data). The more rigorous independent benchmark is the MIT/Stanford close-cycle study (n=277, peer-reviewed): 7.5 days removed from monthly close, 55% more clients per accountant.
  • 50% of finance leaders plan multi-agent AI orchestration across workflows — a meaningfully higher risk profile than single-tool deployments. Multi-agent systems handling invoice approval, GL updates, and vendor payments without human checkpoints require governance architecture mid-market teams typically have not built.
  • The primary deployment barrier is not cost or technology: 64% cite lack of role-specific use cases and 61% cite lack of hands-on practice environments. Finance teams know what AI can theoretically do; they do not know what to do with it in their actual workflow tomorrow.
  • 48% are concerned about accuracy of AI-generated financial outputs — the one concern specific to the function’s core professional obligation. The deployment pattern that handles this: AI for draft generation and anomaly detection, with mandatory human review on all external-facing or regulatory-filing outputs.
  • CFO implication: the 93% figure is a planning input, not a peer-pressure argument. Your finance function will be deploying or scaling AI in the next 18 months by deliberate design or by default — through embedded AI features existing vendors are activating regardless. Build the governance architecture before the multi-agent wave arrives.

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

Grant Thornton “2026 AI Impact Survey” — Governance Readiness as a Budget Prerequisite (n=950, Feb–Mar 2026)

78% of organizations lack confidence they could pass an independent AI governance audit within 90 days. 46% of executives name governance and compliance failures as the leading cause of AI underperformance. Only organizations that have built governance infrastructure (the 22% that are “fully integrated”) report the 58% revenue-growth rate vs. 15% for those still piloting. The CFO implication: the governance infrastructure is not optional overhead on the AI program budget — it is what separates the 15% from the 58%.

Budget checkpoint: before approving the next AI business case, require a governance-readiness line item covering policy documentation, incident response testing, and audit-trail infrastructure. If that line is missing, the program is in the 78% that cannot pass a 90-day audit review.

Source: research/04-consulting-firms/grant-thornton-ai-impact-survey-2026.md — TIER 1, n=950, MEDIUM-HIGH credibility

Supporting research

What this means for mid-market buyers

  • Build a three-year cost model before you sign a vendor agreement. The Year Zero foundation spend ($75K–$175K for a 500-person firm) is the cheapest risk mitigation available — skipping it produces the 380% production-scale overrun MIT Sloan documents.
  • Use BCG’s 10/20/70 rule as a budget sanity check: 10% algorithms, 20% technology, 70% people and process. If your proposed AI budget inverts those proportions, the program will underperform.
  • Treat consumption-based pricing as a Year Two problem to solve in Year Zero. Variable costs grow 30–50% year-over-year even without adding users; lock in pricing caps before deployment, not after the shock.
  • Before approving the next AI license purchase, require a workflow-redesign line item in the same budget request. If the accompanying redesign spend is less than 20–30% of the license spend, the investment case is the 59% tech-first pattern Deloitte documents — and the 1.6x ROI shortfall is the expected outcome, not the downside scenario.

CFO’s Own Workflows: Where Finance Function AI Delivers (Pass 564, April 2026)

The CFO is not just the approver of AI spend — the finance function itself is a high-priority automation target. Independent research and practitioner data as of early 2026 identify five workflows ranked by evidence quality:

  • Financial close has the strongest evidence: MIT Sloan and Stanford (Choi & Xie, August 2025, n=277 accountants, 79 SMBs) document 7.5 days cut from the monthly close and 55% more clients per accountant per week. This is peer-reviewed — the only independent dataset on close automation available.
  • AP automation has the best practitioner case studies: Fanatics Betting & Gaming reduced month-end AP from 20 hours to 2 hours (Bain Capital Ventures CFO survey, n=50, Feb 2025). Vendor-published results (Logitech 83% straight-through, Primark 98% invoice match rate) are directionally consistent but uncontrolled.
  • FP&A and variance analysis adoption lags the opportunity. Commentary generation is the most common AI use (57% adoption in the FP&A function) but variance analysis automation sits at only 30% and forecasting at 28% — the highest-impact analytical workflows are the least deployed (Accounting Today FP&A survey, 2025).
  • Board reporting AI is widely deployed for document production (slide assembly, commentary drafting) but shallow on analytical depth. Governance requirements — defined human review before any AI-generated figure reaches the board — are non-optional.
  • Covenant monitoring has the least evidence. AI can shift monitoring from quarterly to daily and from reactive to predictive, but no independent study provides time-reduction data. The business case is risk asymmetry (breach costs vs. monitoring cost), not efficiency.

Only 17% of finance teams actively use AI in core workflows as of March 2026. Finance ranks last among all business functions in AI deployment maturity. The constraint is execution risk in a function where errors carry audit and investor consequences — not skepticism about the technology (87% of CFOs expect AI to be important to finance operations in 2026, Deloitte Q4 2025 CFO Signals, n=200).

Full analysis: research/05-analyst-firms/cfo-ai-workflow-automation-finance-function.md Detailed workflow breakdown: wiki/cfo-ai-workflows.md

IDC FERS Wave 9: IT Leader Spending Intentions at the Moment of Budget Finalization (Nov/Dec 2025, n=1,007)

The FERS series is a longitudinal bi-monthly IDC survey of IT leaders. Wave 9 is distinctive because fieldwork coincided with budget finalization — stated intentions are closer to actual allocations than any post-hoc survey can deliver.

  • Top three 2026 spending drivers, in order: (1) building custom AI agents, (2) modernizing corporate datacenters, (3) migrating applications to cloud. The ordering matters: agents are the anchor, infrastructure follows.
  • 48% of organizations are prioritizing investment in customized AI agents to automate business processes — this is not a “someday” category for the plurality of the sample; it is a funded 2026 line item.
  • 80.1% of organizations believe agentic AI investment will eliminate manual and semi-manual workflows. This is a belief figure, not a performance figure — the gap between belief and current execution (3% are scaling agents today, per the IDC Agentic AI Adoption Study, n=900+) is the budget-planning risk.
  • Wave 10 (January 2026) direct follow-on: AI/agent governance and security is now funded at 16.7% of total planned AI investment — near parity with infrastructure and applications. This is the first FERS wave to show governance reaching this share; it confirms organizations have internalized that governance must be budgeted before deployment, not after.
  • Infrastructure caveat for CFOs: Agentic workloads consume 10–20x more compute per task than assistive AI (Deloitte, Benchmarkit). An organization that approved AI assistant budgets and then expanded to agents absorbed the multiplier without reforecasting. The datacenter modernization spending in Wave 9 is partially the reckoning for that gap.

Source: research/05-analyst-firms/idc-fers-wave9-ai-spending-2026.md · Nov/Dec 2025 fieldwork · MEDIUM-HIGH · TIER 1

Oliver Wyman Forum CEO Agenda 2026 — What the ~25% Zero-Revenue Cohort Tells CFOs (Apr 2026, n=415 CEOs)

The Oliver Wyman survey is the only large-n CEO-specific AI ROI dataset in this corpus and provides the clearest executive-level signal on where AI budget is failing to generate financial returns.

  • ~25% of CEOs report zero revenue impact despite moving past pilots — concentrated in companies that deployed AI onto existing workflows without redesigning them. The CFO implication: tool deployment without workflow redesign is the mechanism generating this quarter of the sample’s zero-impact result.
  • 53% say it is too early to assess (up from 41%) — capital is deploying faster than accounting systems can measure it. CFOs need measurement infrastructure (token-cost attribution, output quality tracking) to surface returns before the next budget cycle.
  • Deployment leaders redesign workflows at 49% vs. 38% average — the 11-point gap in CEO-level organizational design decisions maps to the BCG 10/20/70 budget allocation rule (10% technology, 20% process, 70% people/change). CFOs approving AI spend without allocating budget to the 70% are systematically underfunding the drivers of return.
  • 43% of CEOs are deprioritizing junior hiring (up from 17%) — workforce cost projections built on this expectation should be stress-tested against the deployment-leader counter-signal: 24% of leaders plan to increase junior hiring, treating AI as an amplifier rather than a substitute.

Source: research/04-consulting-firms/oliver-wyman-ceo-agenda-2026.md

Goldman Sachs Economic Research — The Macro Verdict CFOs Need (Feb–May 2026)

Goldman Sachs Chief Economist Jan Hatzius provides the most independent large-scale assessment of whether AI investment is generating measurable returns.

  • “Basically zero” GDP contribution from AI investment in 2025 — hardware imports offset domestic AI investment contributions, leaving no net measurable output gain at the macro level. CFOs should treat macroeconomic AI ROI narratives with appropriate skepticism until the micro evidence catches up.
  • 30% productivity improvement in two specific use cases (customer service and software development with deep workflow integration) — the gap between zero and 30% is entirely explained by whether AI is embedded in the work process vs. layered on top of it
  • S&P 500 earnings call analysis: 70% of companies discuss AI on calls; the financial statements do not yet corroborate the narrative. CFOs building board decks should note this divergence when using competitor AI investment as a benchmark for their own spending
  • Goldman’s counter-interest credibility: Goldman has significant investment banking relationships with AI sector companies, creating incentive toward optimism. The cautious-to-bearish macro findings run against that interest — which adds credibility to the assessment.

Source: research/01-ai-native-landscape/goldman-sachs-ai-economic-research-2026.md · Feb–May 2026 · HIGH · TIER 1

Foundry State of the CIO + AI Priorities Study 2026 — Where the Budget Is Going and What’s Blocking Returns (n=911 + n=538, 2026)

The Foundry dual survey is the clearest CIO-specific budget dataset in the 2026 corpus — tracking dedicated AI budget adoption across three consecutive years, with deployment blocker detail that no C-suite survey provides.

  • 65% of IT leaders now have a dedicated AI budget (up from 49% one year ago, 36% two years ago — +29pp over two years). This trajectory is the single clearest signal that AI has crossed from “experimental allocation” to “managed program” status in most enterprises.
  • AI is the #1 area of planned IT spending increase (71% of IT leaders) — outranking cloud, security, and ERP modernization in budget priority.
  • 31% still lack a clear corporate AI strategy despite the budget growth — the accountability gap between “funding the program” and “knowing what the program must produce” is where ROI disappears. Cross-reference: Davenport Return on AI Institute (n=1,006) found formal AI strategy triples ROI likelihood.
  • 33% cannot determine AI ROI and 24% are uncertain which department owns AI goals.
  • Infrastructure, skills, and measurement tools are the top investment categories — not new AI models or copilot licenses. This sequence matches Gartner’s 4x investment differential finding and the BCG 10/20/70 budget allocation rule.
  • 67% of IT leaders prefer industry-specific AI vendors over generic horizontal platforms — a vendor selection signal with direct budget implications.

Source: research/05-analyst-firms/foundry-state-of-cio-ai-priorities-2026.md · 2026 · MEDIUM-HIGH · TIER 1

Grant Thornton 2026 AI Impact Survey (n=950, Feb–Mar 2026)

  • 83% of finance functions are increasing 2026 AI budgets — but the revenue return bifurcates sharply: organizations with fully integrated AI report AI-driven revenue growth at 58%, versus 15% at organizations still piloting. Increasing budget without closing the governance gap moves firms into the 15% cohort.
  • 51% of executives name strategy as the top AI ROI driver, but only 22% of operations leaders have a fully implemented strategy — the gap between knowing the lever and pulling it explains the 58/15 revenue split more directly than any technology factor.
  • 34% of finance leaders say AI training is underfunded — a direct budget allocation signal. The survey frames this as a CFO decision: capital is going to models and licenses, not to the change management and workforce capability that determines whether models produce returns.
  • Organizations that pass a governance audit are 4x more likely to report revenue growth. This makes governance investment a CFO decision with a measurable return distribution, not a compliance cost center.

Source: research/06-security-frontier/grant-thornton-ai-impact-survey-2026.md · Grant Thornton n=950 Feb–Mar 2026 · MEDIUM-HIGH · TIER 1

Evanta / Gartner C-Suite Leadership Perspectives 2026 (n=2,505, March 2026)

Largest concurrent cross-functional C-suite AI-spending snapshot in the 2026 corpus: three parallel Gartner surveys covering CIO (n=990), CHRO (n=430), CISO (n=1,085). TIER 1.

  • 66% of CIOs are investing in AI/ML — the single largest planned spending category, ahead of cybersecurity (45%) and data/analytics (50%+). For CFO budget modeling, this establishes AI infrastructure as the dominant CIO line item in 2026, not a discretionary experiment.
  • 50% of CHROs are investing in AI solutions, despite 43% citing rising operational costs as their primary concern — confirming that workforce AI investment is a response to deployment pressure, not a discretionary budget item. The investment is happening even under cost constraints.
  • 43% of CISOs are investing in AI products and services, making it their #1 spending category above DLP and IAM. Security AI investment is accelerating faster than the security function’s traditional spending priorities — relevant for CFO conversations about security budget allocation.
  • The cross-functional investment pattern (66% CIO / 50% CHRO / 43% CISO) establishes a baseline for CFO budget calibration: AI spending has moved from IT to a multi-function commitment. Organizations spending below these share levels in 2026 are materially behind peer investment pace.

Source: research/05-analyst-firms/evanta-gartner-clevel-leadership-perspectives-2026.md · Evanta/Gartner, n=2,505, March 2026 · MEDIUM-HIGH · TIER 1

Futurum Group 1H 2026 Enterprise Software Decision Maker Survey (n=830, early 2026)

Enterprise AI ROI measurement crossed a structural threshold in 2026: direct financial impact (revenue growth + profitability) nearly doubled year-over-year to 21.7% as the primary success metric, while productivity gains fell 5.8pp to 18.0%.

  • Agentic AI surged 31.5% YoY as a top technology priority — the fastest-growing category in the survey, signaling that the pilot era is closing and budget is moving toward autonomous workflow automation.
  • Platform consolidation is accelerating: 65.9% of enterprises run on integrated platforms (up from 60.0%), and 41.0% are actively reducing application count — budget shifting from tool sprawl to stack consolidation.
  • CFO measurement shift: the transition from “productivity hours saved” to “revenue and profitability impact” as the primary AI ROI metric is a leading indicator that AI budget decisions are moving from CTO/CISO to CFO ownership.
  • Caution: Futurum’s business model is research subscriptions sold to technology vendors; YoY comparability is limited by a methodology change in how financial performance metrics were split. Cross-reference Grant Thornton (n=950) and Fed CFO Survey for independent validation of the metric-shift direction.

Sources: research/01-ai-native-landscape/futurum-enterprise-ai-roi-1h-2026.md · research/09-ai-adoption-cycle/futurum-enterprise-ai-roi-shift-2026.md · Futurum Group n=830 · MEDIUM-HIGH · TIER 1

NBER WP35046: What Expert Economists Actually Forecast for AI’s Economic Impact (April 2026)

The authoritative answer to the CFO board question “when will AI show up in the numbers?” — from economists best positioned to assess it.

  • 61.4% of specialist economists expect moderate or rapid AI progress by 2030, yet their unconditional GDP forecast is just 2.5% annualized — barely above government baselines. Capability optimism and economic optimism are not the same forecast.
  • Diffusion lag is the #1 cited constraint. Written rationales drew direct analogies to electrification, automobiles, and PCs — technologies where multi-decade lags routinely separated capability arrival from measurable productivity impact. Energy, chips, and data center bottlenecks are the physical manifestation of this lag.
  • For CFO budget planning: “AI is progressing fast” and “AI will lift our productivity numbers this year” are independent claims. The first has 61% expert support. The second requires a separate organizational adoption assessment. Most CFOs conflate them; this study provides the separation.
  • Rapid scenario GDP by 2045-2049: 3.5% (economists) — “not historically unprecedented,” comparable to post-WWII growth. The floor on rapid-AI upside is not dramatically higher than historical baseline.
  • Inequality risk as demand constraint: Top-10% wealth share forecast to rise from 71.2% (2023) to ~80% by 2050 under rapid AI — if gains concentrate at the top, aggregate demand constraints suppress the macro growth that would otherwise appear in company revenues.

Source: research/01-ai-native-landscape/nber-karger-forecasting-economic-effects-ai-2026.md · NBER WP35046, Karger/Kuusela/Abaluck et al., April 2026 · HIGH · TIER 1

IBM IBV AI Agents Essential — Budget Velocity and IT Spend Shift (n=2,900/2,500, June 2025)

The largest executive survey on AI agent adoption signals an imminent budget composition shift that CFOs should model now.

  • AI investment projected to double as a share of IT budgets: from 12% in 2024 to 20% by 2026 — a near-doubling in two years; the CFO who has not updated the AI budget envelope is already behind peers
  • 64% of current AI budgets are allocated to core business functions, not experiments — the cost-center framing of AI spend has structurally shifted to revenue-adjacent investment in most large enterprises
  • The three blocking factors that prevent AI-first ROI: data readiness (49%), trust deficits (46%), skills gaps (42%) — none resolve by purchasing more software; all require dedicated budget line items outside the AI tool budget
  • ⚠️ TIER 2 SOURCE: Published June 2025 — fieldwork predates current model generation. Agent capabilities have advanced materially; budget forecasts may understate current velocity

Source: research/12-agent-workers/ibm-ibv-ai-agents-essential-survey-2025.md · IBM IBV / Oxford Economics, n=2,900, June 2025 · MEDIUM · TIER 2

IBM IBV Enterprise 2030 AI Ambition Gap — The 79/24 Problem (n=2,007, Q3–Q4 2025)

The widest gap in corporate AI budgeting: most executives expect AI-driven revenue by 2030 but cannot build a budget case for it.

  • 79% of senior executives expect AI to significantly contribute to revenue by 2030 — only 24% can articulate where that revenue will come from — the gap is the CFO’s primary planning problem: how do you fund an ambition you cannot model?
  • 68% of executives fear their AI efforts will fail due to lack of integration with core business activities — not lack of technology; the budget failure is connective tissue, not tooling
  • AI investment projected to surge 150% as a share of revenue between 2025 and 2030 — executives who have not built a multi-year AI budget envelope are already behind the planning curve
  • Multi-workflow AI adopters anticipate 24% greater productivity and 55% higher operating margins — the budget decision is not whether to invest; it is whether to invest with integration scope or tool scope

Source: research/07-adoption-challenges/ibm-ibv-enterprise-2030-ai-ambition-gap-2026.md · IBM IBV / Oxford Economics, n=2,007, Q3–Q4 2025 · MEDIUM-HIGH · TIER 1

Gartner Worldwide AI Spending Forecast 2026 — The Services Premium (Jan 2026)

Market-level spending data that validates and contextualizes enterprise-level AI budget architecture decisions.

  • $2.52 trillion worldwide AI spend (+44% YoY from $1.76T in 2025) — the industry aggregate that anchors benchmarking: a mid-market organization spending 0.5% of revenue on AI is below the global intensity curve
  • Services ($589B) > software ($453B) by $136B — the most important budget ratio: for every dollar in AI software licensing, budget $1.30 in services; organizations using software-only budget models systematically underfund deployment
  • AI models ($26.4B) = 1% of total spend — negotiating hard on API/model pricing optimizes a rounding error; implementation, integration, and change management are the dominant costs
  • Trough of disillusionment framing — Gartner’s official Hype Cycle placement for 2026: AI will be bundled by incumbent vendors rather than pitched as standalone; expect consolidation pressure on point solutions
  • 2027 projection: $3.34 trillion (+32% from 2026) — growth rate decelerates as market matures; CFOs modeling multi-year AI budgets should use this curve, not 44% YoY in perpetuity

Source: research/05-analyst-firms/gartner-worldwide-ai-spending-forecast-2026.md · Gartner IT Spending Forecast, Jan 15, 2026 · MEDIUM-HIGH · TIER 1

IBM IBV: ERP as AI Scaling Vector — ROI and Margin Data (Jan–Mar 2025, n=1,500)

ERP-platform AI embed as the primary CFO ROI lever for mid-market organizations already on SAP.

  • 27% higher ROI and 9% stronger operating margins in organizations that aggressively embed AI into ERP vs. cautious adopters — organizational capabilities (skills, governance, training) drive the gap, not investment level
  • 76% of “AI Bullish” executives fund AI at high levels vs. 58% of “AI Bearish” — but the marginal ROI driver is capability readiness, not spend; organizations that fund without building governance produce the cautious-adopter outcome regardless of budget
  • 4.4x integration multiplier: organizations that embed AI across multiple ERP modules and business functions report 4.4x the ROI of single-module deployments — the budget implication is that integration scope is an ROI lever, not merely an implementation cost
  • Top 3 CFO-relevant capability gaps: workforce skills (the largest), governance/policy readiness, and cross-functional AI training tied to real workflows — the same pattern independently corroborated by BCG (n=13,000), McKinsey (n=1,993), and Gartner (n=500+)

Apply vendor caveat: IBM IBV surveys only SAP users; IBM has direct commercial interest in ERP-centric AI consulting. Directional findings consistent with independent research; magnitude claims (27% ROI, 4.4x) are correlational and self-reported with no control group.

Source: research/04-consulting-firms/ibm-ibv-erp-meets-ai-2026.md · IBM IBV, n=1,500, Jan–Mar 2025 · MEDIUM · TIER 2

Gartner CIO Agenda 2026 — Budget Intent vs. Deployment Reality (n=2,501)

The world’s largest CIO survey quantifies the tension between AI spending commitment and operational execution:

  • 91% of CIOs are increasing GenAI budgets in 2026, mean increase +38% — the highest investment growth rate across all technology categories tracked
  • Only 17% have deployed AI agents; 64% plan to within 24 months — budget is running well ahead of deployment
  • Only 48% of digital initiatives currently meet or exceed business targets — the success baseline before AI complexity is added
  • Strategic insourcing via AI can reduce operational costs 5–30%, but the range collapses or expands almost entirely based on workflow redesign quality, not budget size
  • On-premises infrastructure is the only category where budgets are shrinking (41% decreasing, mean −5%) — cloud and software absorb AI spend while hardware decommissioning lags

The CFO implication: a +38% AI budget increase directed toward a category where only 17% of organizations have operational deployments and 48% of digital initiatives meet targets is a funding posture that requires governance checkpoints, not just quarterly budget reviews.

Source: research/05-analyst-firms/gartner-cio-agenda-2026.md — MEDIUM-HIGH / TIER 1 (n=2,501, Jan 2026)

See also

  • CFO AI Workflows — the five finance function workflows ranked by evidence quality and implementation readiness
  • Inference Economics — unit economics that shape consumption-based pricing
  • AI Vendor Contracts — contract terms that control consumption surcharges
  • Data Readiness — the Year Zero investment that determines Year One and Year Two costs
  • Board AI Strategy — how the 3-year cost model surfaces at the board level

Backlink: research/01-ai-native-landscape/ai-budget-benchmarking-mid-market.md — mid-market AI spending benchmarks: Year 0–2 cost curves, 250–999 employee company spending at 8.4% of IT budget, RSM/Deloitte/BCG synthesis — MEDIUM-HIGH / TIER 1–2