← Adoption Challenges 12 min read

AI and Capital Allocation: The Portfolio Question Every CFO Must Answer Before Writing the First Check

Brandon Sneider | March 2026

Executive Summary

  • AI now commands 8-13% of enterprise IT budgets and rising — but 86% of CFOs have not seen measurable returns. RGP’s survey (n=200 U.S. finance chiefs, October-November 2025) finds only 14% report clear, measurable impact from AI investments. Yet 91% plan to increase spending again this year (Deloitte, n=1,854, August-September 2025). The gap between investment momentum and realized value is the central capital allocation problem in American business.
  • AI is not competing for new money — it is winning a zero-sum fight against headcount, ERP modernization, and infrastructure. Gartner (n=300+ CFOs, October 2025) finds headcount growth expectations collapsing from 6% to 2%, HR budgets cut to 0.7% growth, and technology budgets rising for 75% of CFOs. Every AI dollar comes from somewhere, and the companies that win treat this as portfolio management, not a technology decision.
  • The 6% that capture meaningful EBIT impact allocate more than 20% of digital budgets to AI — triple the median. McKinsey (n=1,993, June-July 2025) identifies this threshold as the single clearest differentiator between high performers and the rest. Below it, AI investment delivers activity without impact.
  • The capital allocation decision is a sequencing problem, not a sizing problem. The evidence points to a consistent pattern among high performers: fund one workflow deeply before spreading capital across many, invest 70% in people and process redesign before buying more technology, and establish kill criteria before writing the check — not after the board starts asking questions.
  • CFOs who treat AI as a portfolio bet — not a line item — outperform by 2x on ROI. Deloitte finds companies that redesign work alongside AI deployment are twice as likely to exceed ROI expectations. PwC puts it bluntly: technology delivers about 20% of an initiative’s value; the other 80% comes from redesigning work.

The Portfolio Problem: What the CFO Actually Faces

A mid-market CFO in 2026 faces a capital allocation landscape unlike anything in the prior decade. Five investment categories compete for the same budget:

Investment Category Typical Mid-Market Claim Annual Budget Range (400-Person Co.) Risk Profile
AI programs (new) 8-13% of IT budget, growing 38% YoY $75K-$500K High variance: 14% see returns, 86% do not
ERP modernization 6-15% of tech budget, multi-year $200K-$2M Predictable but slow: 18-36 month payback
Headcount expansion Historically 6% growth, now 2% $85K-$150K per hire (fully loaded) Immediate capability, ongoing obligation
Cybersecurity Protected category, growing 26% $150K-$500K Regulatory mandate, not discretionary
Infrastructure/cloud Consolidation phase, AI-dependent $100K-$400K Foundation for all other investments

The CFO’s dilemma is that AI demands capital now but delivers returns over 2-4 years (Deloitte, n=1,854, August-September 2025), while ERP and headcount have established — if imperfect — ROI timelines. The temptation is to fund AI incrementally from discretionary budgets. The evidence says this approach fails.

Why Incremental AI Funding Produces Incremental Results

The data on AI investment thresholds is converging across sources.

The 5% threshold. Deloitte’s tech executive survey (n=548, May-June 2025) finds organizations that allocate more than 5% of their IT budget to AI see 70-75% of projects yield positive results, versus 50-55% for those below the threshold. The marginal return on the investment dollar that crosses this line is disproportionately high.

The 20% differentiator. McKinsey (n=1,993, June-July 2025) finds AI high performers — the 6% that report 5%+ EBIT impact — allocate more than 20% of digital budgets to AI. Other organizations allocate roughly 7%. The gap is not just magnitude: high performers are 2.8x more likely to have redesigned workflows, 5x more likely to have dedicated AI teams, and far more likely to have moved from pilot to production at scale.

The critical mass problem. Grant Thornton’s Q1 2026 CFO survey (n=230+ finance leaders, Q1 2026) confirms that 68% of CFOs expect IT spending increases — the highest in 21 quarters. But spreading those increases thinly across AI, ERP, security, and infrastructure means no category reaches the investment density required to generate returns. Deloitte’s State of AI (n=3,235, August-September 2025) finds 37% of organizations still use AI “at a surface level with no process change.” That 37% is, in large part, the result of capital allocation that funds experimentation but not transformation.

The pattern is clear: AI capital allocation that stays below threshold produces the worst possible outcome — cost without returns and organizational fatigue without organizational learning.

The Five Trade-Offs: What Gets Funded, What Gets Deferred

Trade-Off 1: AI vs. Headcount

This is the largest and most consequential reallocation.

Gartner’s data (n=300+ CFOs, October 2025) is unambiguous: headcount growth expectations have collapsed from 6% to 2%. Only 21% of CFOs plan staff increases of 4-9%, down from 31% the prior year. A Fortune survey (n=350+ public-company CEOs and investors, March 2026) puts it more starkly: 66% plan to freeze or cut hiring through 2026. Entry-level listings have dropped 30%, middle management postings 42%.

The trade-off calculus: One deferred hire at $85K fully loaded funds a meaningful AI pilot. Two deferred hires fund a Year 0 program. But Gartner separately warns (March 2025 press release) that CFOs should “reset expectations about AI’s impact on workforce productivity and headcount” — because 42% anticipate AI-driven headcount reductions across SG&A, yet only 33% expect reductions as modest as 1-5%. The risk of cutting people to fund tools that do not yet deliver people-level capability is real.

The high-performer approach: Fund AI from deferred hiring, but protect the roles that make AI work — the workflow redesigners, the internal champions, the process owners. BCG’s 10-20-70 framework prescribes 70% of AI investment going to people and processes. Cutting people to fund technology is the precise inversion that produces the 75% failure rate.

Trade-Off 2: AI vs. ERP Modernization

Most mid-market companies allocate 6-15% of their technology budgets to ERP upgrades (industry composite, 2025-2026 surveys). ERP modernization is already underway at the majority. The question is whether AI investment accelerates, delays, or replaces the ERP roadmap.

The convergence reality: The largest ERP vendors — SAP (Joule AI), Oracle (AI in Fusion), Microsoft (Dynamics 365 Copilot) — are embedding AI directly into their platforms. For companies already committed to a major ERP vendor, the AI investment and the ERP investment are converging. The capital allocation question becomes not “AI or ERP” but “which ERP modules get AI-augmented first.”

The mid-market trap: Companies that defer ERP modernization to fund standalone AI tools often discover that their AI tools cannot integrate with their legacy systems. RGP finds 86% of CFOs identify legacy tools as a “significant or moderate barrier” to AI adoption. The capital saved by deferring ERP modernization may be spent twice — once on AI tools that cannot connect, and again on the deferred ERP upgrade that enables the connection.

The decision rule: If the ERP vendor’s AI capabilities cover the target workflow, fund AI within the ERP budget. If the target workflow sits outside the ERP footprint (customer service, marketing, legal research), fund AI separately. Do not defer ERP modernization that AI tools depend on.

Trade-Off 3: AI vs. Cybersecurity

This is not a real trade-off. Cybersecurity spending is a protected category in 2026, growing 26% (Gartner, 2025). AI deployment actually increases the cybersecurity budget requirement — data classification for AI tools, prompt injection defenses, model access controls, and vendor security reviews are new line items that did not exist three years ago.

The capital allocation implication: Budget AI security costs as part of the AI investment, not the cybersecurity budget. The AI program owns the incremental risk. Failing to account for this is how companies underestimate total cost of ownership by 40-60% (Mavvrik, 2025).

Trade-Off 4: AI vs. Infrastructure/Cloud

Gartner’s CIO survey (n=2,501, May-June 2025) shows on-premise infrastructure as the only budget category with negative spending intent (-5%). Cloud spending grows 21%. Data center spending surges 31.7%.

The reallocation pattern: Infrastructure budgets are not disappearing — they are being redirected from maintaining legacy on-premise systems to building AI-capable cloud foundations. For mid-market companies, this means the infrastructure conversation is inseparable from the AI conversation. The AI program needs data infrastructure to function; the data infrastructure investment needs an AI use case to justify itself.

Trade-Off 5: AI vs. SaaS Portfolio

SaaS waste is the easiest funding source and the least politically difficult trade-off. License utilization sits at 54% across enterprises (Zylo, 40M+ licenses, 2026). For a mid-market company spending $4,830 per employee on SaaS, roughly $2,200 per employee per year is waste. A 400-person company recovering half of that waste frees $440K — enough to fund a Year 1 AI program without cutting a single other category.

West Monroe’s survey (n=310 executives, 2025) confirms the opportunity: nearly half of organizations saw licensing and subscription costs increase more than the industry average of 10%. AI itself is contributing to the inflation — vendors embedding AI features into existing products and raising prices 10-30%.

The action step: Run a SaaS rationalization before budgeting AI. The waste found in unused licenses, redundant tools, and underutilized subscriptions frequently covers the Year 0 AI investment without touching headcount, ERP, or infrastructure budgets.

The Portfolio Framework: How High Performers Allocate

Synthesizing the data across McKinsey, Deloitte, BCG, and Gartner, the companies that capture measurable AI value share a capital allocation discipline with five characteristics:

1. Concentrate before you diversify. PwC’s 2026 predictions emphasize that “crowdsourcing AI efforts can create impressive adoption numbers, but it seldom produces meaningful business outcomes.” High performers pick one workflow, fund it to the 5% IT budget threshold, and prove value before expanding. Scattering $50K across five departments is five experiments; investing $250K in one workflow is a transformation.

2. Fund the 70%, not just the 30%. BCG’s 10-20-70 framework prescribes 70% of AI investment in people and processes, 20% in technology, 10% in algorithms. Most companies invert this ratio, spending 60-70% on technology. The capital allocation decision is not “how much for AI tools” — it is “how much for the training, workflow redesign, and change management that make AI tools produce returns.”

3. Establish kill criteria before funding. Pre-approval success metrics produce a 54% project success rate versus 12% without them (Pertama Partners, n=2,400+ AI initiatives, 2025-2026). The capital allocation decision should include a 90-day checkpoint with defined gates: adoption below 25% triggers review, cost per transaction rising triggers investigation, no P&L impact at 12 months triggers shutdown.

4. Sequence AI with infrastructure, not against it. Companies that fund AI before fixing data infrastructure spend the money twice. The 35% of CFOs who cite data trust as their top barrier to AI ROI (RGP, n=200, October-November 2025) are experiencing this sequencing failure in real time. The capital allocation sequence is: data readiness first, AI deployment second.

5. Rebalance quarterly, not annually. AI investment economics change faster than annual budget cycles. The HBR/NewVantage survey (n=100+ Fortune 1000 executives, 2026) finds 39% now have AI in production at scale, up from 5% two years prior. Companies that lock AI budgets into annual cycles miss the opportunity to redirect capital from underperforming pilots to proven workflows.

Key Data Points

Metric Finding Source
CFOs seeing measurable AI ROI 14% RGP (n=200, Oct-Nov 2025)
Organizations increasing AI investment 91% Deloitte (n=1,854, Aug-Sep 2025)
AI high performers’ digital budget share >20% McKinsey (n=1,993, Jun-Jul 2025)
Other organizations’ digital budget share ~7% McKinsey (n=1,993, Jun-Jul 2025)
Organizations reporting meaningful EBIT impact 6% McKinsey (n=1,993, Jun-Jul 2025)
Headcount growth expectations (2025 vs. 2026) 6% → 2% Gartner (n=300+, Oct 2025)
CFOs planning tech budget increases 75% Gartner (n=300+, Oct 2025)
IT budget threshold for AI project success >5% Deloitte (n=548, May-Jun 2025)
SaaS license utilization rate 54% Zylo (40M+ licenses, 2026)
Legacy systems as barrier to AI ROI 86% RGP (n=200, Oct-Nov 2025)
ROI improvement from work redesign 2x Deloitte (n=1,854, Aug-Sep 2025)
AI’s share of initiative value (technology alone) ~20% PwC (2026 predictions)
Pre-approval success criteria impact 54% vs. 12% Pertama Partners (n=2,400+, 2025-2026)

What This Means for Your Organization

The capital allocation question is not “can I afford AI?” It is “can I afford to fund AI the way 86% of companies do — incrementally, without workflow redesign, and without kill criteria — and expect to be in the 14% that see returns?”

The evidence is consistent: AI capital allocation is a portfolio management problem, not a technology procurement problem. The CFO who treats the AI budget as a line item within the technology budget has already made the mistake. The AI investment touches headcount planning, ERP roadmaps, SaaS rationalization, and infrastructure modernization simultaneously. It demands a portfolio view.

For a 400-person company considering its first meaningful AI investment, the practical sequence is: rationalize SaaS waste first (the $200K-$440K found there funds Year 0 without touching other budgets), invest 70% of that in workflow redesign and training (not tools), establish 90-day kill criteria before deployment, and protect the infrastructure modernization that AI depends on. The companies that reverse this sequence — buying tools first, figuring out processes second, hoping data infrastructure catches up — are the ones funding the 86% failure rate.

If this framework raised questions about how your specific capital allocation picture should shift, I would welcome that conversation — brandon@brandonsneider.com.

Sources

  1. RGP, “The AI Foundational Divide: From Ambition to Readiness” (n=200 U.S. CFOs, October-November 2025, published December 2025). Independent advisory firm survey. Companies $500M-$10B+ revenue. High credibility for CFO-specific data. https://rgp.com/press/rgp-cfo-survey-shows-growing-divide-between-ai-ambition-and-ai-readiness/

  2. Deloitte, “AI ROI: The Paradox of Rising Investment and Elusive Returns” (n=1,854 senior executives, 14 countries, August-September 2025). Large sample, cross-industry. Note: skews to larger enterprises, European-heavy sample. https://www.deloitte.com/global/en/issues/generative-ai/ai-roi-the-paradox-of-rising-investment-and-elusive-returns.html

  3. McKinsey, “The State of AI in 2025” (n=1,993, June-July 2025). Gold-standard annual survey. The 6% EBIT impact threshold for “high performers” is the most actionable finding. Note: 38% of respondents from $1B+ companies — mid-market is undersampled. https://www.mckinsey.com/capabilities/quantumblack/our-insights/the-state-of-ai

  4. Gartner, “CFOs’ Budget Plans Prioritize Growth Functions, Technology and AI in 2026” (n=300+ CFOs, October 2025, published February 2026). Authoritative for budget intent data. Note: minimum $1B revenue for respondents; mid-market extrapolation required. https://www.gartner.com/en/newsroom/press-releases/2026-02-10-gartner-research-reveals-cfos-budget-plans-prioritize-grotwth-functions-tech-and-ai-in-2026

  5. Grant Thornton, “CFOs Accelerate Tech Spending as AI Momentum Increases” (n=230+ finance leaders, Q1 2026). Mid-market-friendly sample. 68% expecting IT spending increases is the 21-quarter high. https://www.grantthornton.com/insights/press-releases/2026/march/cfos-accelerate-tech-spending-as-ai-momentum-increases

  6. Deloitte, “Q4 2025 CFO Signals Survey” (n=200 North American finance chiefs, November-December 2025). 87% call AI “extremely or very important” to finance operations. 54% plan AI agent integration as transformation priority. https://www.deloitte.com/us/en/about/press-room/deloitte-q4-2025-cfo-signals-survey.html

  7. Deloitte, “State of AI in the Enterprise 2026” (n=3,235, 24 countries, August-September 2025). Largest AI adoption survey. 37% use AI at surface level, 34% report deep transformation, 25% moved 40%+ of pilots to production. https://www.deloitte.com/us/en/about/press-room/state-of-ai-report-2026.html

  8. L.E.K. Consulting, “2025 Office of the CFO Survey” (n=~100 CFOs, 2025). Smaller sample but CFO-specific. 54% believe delaying AI adoption slows growth. 86% prefer embedded AI within platforms. https://www.lek.com/insights/hea/us/ei/lek-consultings-2025-office-cfo-survey-study-ai-ocfo

  9. PwC, “2026 AI Business Predictions” (2026). Framework-oriented, not survey-based. The “20% technology / 80% work redesign” value split is the headline finding. https://www.pwc.com/us/en/tech-effect/ai-analytics/ai-predictions.html

  10. West Monroe, “State of B2B Software & AI Spend” (n=310 executives, 2025). AI at 12-15% of IT budgets. SaaS cost inflation from AI feature embedding. https://www.westmonroe.com/insights/state-enterprise-ai-spend

  11. HBR/NewVantage Partners, “How Executives Are Thinking About AI in 2026” (n=100+ Fortune 1000 executives, 2026). 54% report high business value from AI (up from 47%). 93% cite culture, not technology, as key challenge. https://hbr.org/2026/01/hb-how-executives-are-thinking-about-ai-heading-into-2026

  12. BCG, “AI Radar” survey (n=1,803 C-level executives, January 2025). AI leaders achieve 2x revenue growth and 40% more cost savings. 10-20-70 framework allocation. Consulting firm survey — rate accordingly.

  13. Pertama Partners analysis (n=2,400+ enterprise AI initiatives, 2025-2026). Pre-approval success criteria produce 54% vs. 12% success rates. Median time to project abandonment is 11 months.

  14. Zylo, “SaaS Management Index” (40M+ licenses analyzed, 2026). License utilization at 54%. $4,830 per employee average SaaS spend. Vendor data — rate accordingly but consistent with Gartner’s own estimates.

Brandon Sneider | brandon@brandonsneider.com March 2026