Two-Thirds of Enterprises Have AI Budgets. Only 5% Can Prove They're Working.
Two-thirds of enterprise organizations now have dedicated AI budgets. Ninety-five percent of those AI pilots show no measurable P&L impact. That gap is not a technology failure. It is a measurement and governance failure, and it is costing marketing leaders their credibility with finance.
The numbers are stark. According to Techlevity's analysis of EY and MIT data, 84% of organizations increased AI spend this year, yet only 5% have scaled AI with measurable earnings impact. The Martech Weekly's Enterprise Martech Outlook 2026 puts the martech-specific version of this problem in sharper relief: 40.2% of martech leaders still cannot show a clear financial contribution from their technology investments. We are funding experiments at scale while proving almost nothing.
The Spend-Results Gap Is Structural, Not Technical
The instinct is to blame the technology. The models are immature. The agents hallucinate. The integrations are fragile. But the evidence points elsewhere.
RBD Co.'s cross-industry analysis found that organizations allocate 93% of AI budgets to technology and 7% to the people, processes, and organizational design required to absorb it. BCG estimates that AI success depends approximately 70% on people and processes, 20% on technology infrastructure, and 10% on algorithms. The ratio is inverted. We are buying the engine and skipping the chassis.
This explains why McKinsey's 2026 State of AI survey found that only 37% of respondents attribute any EBIT impact to AI, about the same share as a year earlier. Budgets rose. Overruns got funded. The earnings impact line did not move.
The sunk cost mechanism is doing real damage here. In Futurum's 2H 2026 survey of 1,636 technology decision makers, 46.9% were over their AI budget. Only 17.2% of those paused or reduced an initiative in response. Nearly half asked for more budget instead. We are doubling down on pilots that have not proven anything because we already spent the first tranche.
Token Economics Changed the Cost Model
BCG's analysis of AI budget dynamics surfaces a cost driver most marketing teams have not internalized: token volume. Server prices are running 80% above 2025 levels. Software inflation is four to five times the market rate. Average token costs for top US large language models rose from below $5 per million at the start of 2025 to $20 per million in 2026.
Price is only part of the story. As adoption accelerates, token volume becomes an equally important cost driver. A team of 2,000 people, each saving two hours a week, can also each generate two hours of API costs a week at a rate no one planned for.
One enterprise spent $500 million in a single month on AI services, not because of a breach or a runaway process, but because they had not set spending limits, had not implemented usage controls, and had not built monitoring tooling to detect unusual consumption patterns before the invoice arrived. That is an outlier in scale. The governance failure that produced it is not.
The Winners Are Not Spending More. They Are Spending Differently.
AGI Bible's analysis of the 5% that succeeded found a consistent pattern across XPO, C3.ai, and Upstart: all three locked in a specific metric that already mapped to a financial statement, then asked whether AI could improve that metric. XPO anchored to savings per percentage point of efficiency gained. C3.ai's PANDA platform anchored to a direct comparison of annual operating cost versus annual savings. Upstart anchored to a basis-point spread versus Treasury returns.
All three metrics were numbers finance or operations teams were already tracking. AI was applied to improve an existing, verifiable metric, rather than being treated as a standalone project that needed to invent its own new measurement standard.
This is the reverse order of most AI pilots. Most start with the technology, then search for a use case, then try to invent a measurement framework after the fact. The winners started with the P&L line, then asked whether AI could move it.
The same research found that companies measuring ROI without matching infrastructure investment returned only 8.14%, trailing the benchmark by 2,100 basis points. Measurement surfaces problems. If the company lacks the capacity to fix what measurement reveals, the result is simply a documented admission of failure with no follow-through. The market reads that as a red flag about management capability, not a positive signal.
What This Means for Marketing's AI Business Case
The Martech Weekly's data shows AI is making a bigger dent in marketing operations than in customer experience. Some 69.3% of those surveyed say AI is having a reasonable, clear, or substantial impact on the martech stack. Fewer, 59.9%, say the same about customer experience.
That gap makes sense. Internal uses like writing, editing, analysis, reporting, and creative variation are easier to test, govern, and fit into workflows marketers already know. Co-pilots, copywriting, data management, and data analysis are further along in production and report better ROI. Journey optimization, decisioning, audience selection, campaign creation, and attribution still have more ground to cover.
The implication for marketing leaders is uncomfortable but clear: the AI initiatives easiest to fund are not the ones easiest to prove. The internal productivity gains are real but diffuse. The customer-facing applications that could move pipeline are harder to measure and slower to scale.
The CFO-Safe Path Forward
Stop an AI pilot when you would not approve it today as a new proposal. Price it only on the money still to be spent, and judge it on evidence of a path to EBIT rather than user enthusiasm. Have someone other than the original sponsor make the call.
Georgian's AI Applied benchmarks show that across 18 months of accelerating AI adoption, the share of companies that can name a revenue or cost-savings P&L line tied to their AI program has not moved: Wave 1 at 45%, Wave 2 at 46%, Wave 3 at 45%. The hard ROI measurement gap is the one CFOs will close in on as AI spend becomes a material line item.
The organizations generating returns are not spending more on technology. They are spending differently: anchoring to existing financial metrics, pairing measurement with execution capacity, and treating AI as a budget discipline rather than a technology experiment.
Model or it didn't happen. The board is going to ask. Finance is going to ask. The question is whether you have the spreadsheet ready or whether you are still searching for the metric that proves the spend was worth it.
