84% Are Increasing AI Spend. Only 25% See Transformative Impact.

Excerpt: Companies are spending more on AI than ever before. 95% cannot prove it is working. The gap between investment and return is not a technology problem — it is an economics problem. Here is why more spend does not equal more value, and what to do about it.

Slug: 84-percent-increase-ai-spend-25-percent-see-impact

Categories: ai-strategy, cost-optimisation, leadership

> TL;DR: Despite 84% of organisations increasing AI budgets, only 5% achieve measurable P&L impact from $30-40B in GenAI investment. The problem isn't technology—it's token-based economics creating a spend-results gap. Success requires CFO-level token consumption tracking, outcome-tied budgets, and C-Suite accountability for AI economics.

84% of organisations are increasing their AI spend this year. That is not a forecast. That is a current decision, already made, budgets already allocated. AI investment is accelerating across virtually every sector and geography — yet as our analysis of why AI projects fail shows, spending more doesn't guarantee results.

Yet the returns are stunningly absent. 95% of enterprises report no measurable return on the $30-40 billion collectively invested in generative AI. Ninety-five per cent. Almost every company spending money on AI cannot prove that money is producing results.

Only 25% of organisations describe their AI outcomes as "transformative." The rest are somewhere between "incremental improvement" and "we are not sure." For a technology that dominates every boardroom conversation, that is a remarkably poor conversion rate.

This is not a technology failure. The AI works. Models are more capable than ever. Agents are more autonomous. The tools are real. The failure is in the economics — specifically, in how organisations budget, measure, and manage AI as a financial asset rather than a technical curiosity.

The spend-results gap, in detail

Deloitte's 2026 research on executive decisions shaping digital workforce value, combined with EY and MIT data, paints a precise picture of the disconnect.

Deloitte surveyed 1,854 executives across Europe and the Middle East. Their findings on agentic AI returns are instructive:

• 10% are seeing value now
• 21% expect value within one year
• 29% expect value in one to three years
• 33% expect value in three to five years
• 7% do not expect value for more than five years

Let us read that carefully. Only a third of organisations — 31% — are seeing or expecting to see value from agentic AI within the next year. More than two-thirds are in a waiting pattern, investing now and hoping returns show up eventually.

Meanwhile, only 5% of companies have scaled AI with measurable P&L impact. Five per cent. That is the share of organisations that have moved AI from experiment to enterprise-wide deployment and can point to a specific line on the profit-and-loss statement that changed as a result.

EY and MIT put the same problem in starker terms: 95% of enterprises report no measurable return on $30-40 billion in GenAI investment. That is not a gap. That is a chasm.

Why more spend does not equal more value

The core problem is what Deloitte calls "token-based economics." In traditional IT, you buy a licence and the cost is fixed. A CRM subscription costs the same whether your sales team uses it for ten deals or ten thousand. The marginal cost of usage is zero.

AI does not work that way. Every time an AI model processes a request, it consumes tokens. Every token is both a unit of work and a unit of cost. More usage means more tokens means more cost. The relationship between usage and spend is direct and linear.

This creates a dynamic that most organisations are not prepared for: spending more on AI does not necessarily produce more value, but it always produces more cost. If your AI agents are processing tasks that do not matter, you are spending money on activity rather than outcomes — and that's exactly the gap CIOs must close between AI investment and ROI. The spend goes up. The value does not.

The average planned AI investment across organisations is $186 million. That is a significant line item. And for most companies, it is a black box — money goes in, tokens come out, and nobody can tell you whether the tokens are producing anything worthwhile.

Jevon's Paradox — the trap that guarantees higher bills

Here is where the economics get counterintuitive. Most organisations assume that as AI gets cheaper — as the cost per token decreases — their total AI spend will go down. The opposite is true.

This is Jevon's Paradox, named after the 19th-century economist William Stanley Jevons, who observed that as the efficiency of coal use improved, total coal consumption increased rather than decreased. Cheaper coal meant more applications became economical, which drove total demand up faster than the efficiency savings.

The same dynamic applies to AI tokens. As the cost per token decreases — and it is decreasing rapidly — new use cases become economical. Tasks that were too expensive to automate last year become cost-effective this year. Organisations deploy more agents, process more data, run more workflows. Total token consumption explodes. And total spend goes up, not down — a pattern explained by Amdahl's Law for AI engineering.

Deloitte's research confirms this pattern. As the unit cost of intelligence decreases, the total spending on AI infrastructure increases as digital workforces scale. The savings per unit are real. But the volume increase more than offsets them.

For the CFO trying to budget AI spend, this is critical. If your financial model assumes AI costs will decrease over time, your model is wrong. AI costs will increase — potentially dramatically — as your organisation finds more ways to use the technology. The question is not how to reduce AI spend. It is how to ensure that increasing spend produces increasing value.

The C-Suite responsibility matrix

One reason the spend-results gap persists is that nobody in the C-Suite owns the economics of AI. The CTO owns the technology. The CFO owns the budget. But the question of whether AI spend translates to AI value falls between the cracks.

Deloitte's research assigns specific responsibilities to each C-Suite role:

CEO — AI ambition. The chief executive sets the strategic direction. Is AI a cost reduction play, a revenue growth engine, or a business model transformation? The answer determines how you measure returns.

CFO — Token consumption. Someone in the finance function needs to track, forecast, and optimise token-level spending. If your CFO cannot tell you your monthly token consumption, your AI strategy is a black box — one that AI-native engineering support can help illuminate.

COO — Rethink operations. AI does not optimise existing processes. It replaces them. The COO must identify which operations to redesign, not just which tasks to automate.

CIO/CTO — Infrastructure. The technology leader owns the compute, networking, and platform decisions that determine how much each token costs. Infrastructure efficiency directly impacts token economics.

CHRO — Human-agentic strategy. The HR leader must plan for a workforce that includes both humans and AI agents. This is not a future concern — it is a current planning requirement.

CRO — Governance. The chief risk officer ensures that AI deployment meets regulatory, security, and ethical standards. Governance failures do not just create risk — they destroy value by forcing project cancellations.

Practical steps to close the gap

1. Audit your token economics

Before you increase AI spend, understand where current spend is going. Map your token consumption by use case, by team, by model. Identify which tokens are producing measurable value and which are noise.

2. Tie spend to outcomes

Every AI initiative should have a financial outcome attached before it receives budget. Not "improve efficiency." Not "enhance capability." A specific number: "reduce customer onboarding cost by 20%" or "increase sales conversion by 5 percentage points."

3. Budget for Jevon's Paradox

Your AI budget should assume rising total spend, even as per-unit costs decrease. Plan for it. Allocate contingency. And make sure the rising spend is directed at use cases with clear returns, not at activity that looks productive but produces nothing.

4. Assign C-Suite ownership

Use Deloitte's responsibility matrix to assign clear accountability. If nobody owns AI economics, nobody will optimise them. Start with the CFO — if they cannot articulate your token consumption, that is the first gap to close.

5. Kill the low-value pilots

95% of enterprises see no measurable return. Some portion of that is low-value pilots that will never scale — the exact pattern behind the £200K AI pilot that never shipped. Audit your portfolio of AI initiatives. Kill the ones that have been running for six months without measurable results. Redirect the budget to the few that are working.

Key Takeaways

• Token economics are linear: Unlike traditional IT, AI costs scale directly with usage—more tokens always mean more cost
• Jevon's Paradox applies: Cheaper AI leads to higher total spend as new use cases become economical
• C-Suite accountability gaps: Success requires clear ownership—CEO sets ambition, CFO tracks tokens, COO redesigns operations
• 95% see no ROI: Most AI investments lack measurable outcomes and proper financial tracking
• Kill failing pilots: After 6 months without measurable results, redirect budget to proven use cases

The bottom line

AI spending is increasing. AI returns are not. The gap is not because the technology is broken. It is because the economics are misunderstood. Tokens are not free. Jevon's Paradox guarantees that cheaper AI means higher total bills. And most organisations have no mechanism to connect spend to outcomes.

If your CFO cannot tell you your token consumption, your AI strategy is a credit card with no limit. If your CEO cannot articulate what "transformative" looks like in pounds, you are in the 95% who see no return. The fix is not more technology. It is better economics.

Stop guessing at AI ROI. TechLevity offers a token economics audit that maps your AI spend to your AI outcomes, identifies where value is leaking, and builds a financial model that accounts for Jevon's Paradox. Book yours — your CFO will thank you.

[Book your token economics audit →]

<!-- Internal linking opportunities:

• Link to /services/ai-strategy when discussing strategic AI direction
• Link to /services/fractional-cto for C-Suite leadership gaps
• Link to /services/ai-governance for risk and compliance
• Link to /services/cloud-optimization for infrastructure efficiency
• Link to why-ai-projects-fail article for failure patterns
• Link to ai-sdlc-maturity-framework for development processes
• Link to 200k-ai-pilot-never-shipped for pilot failure cases
• Link to shadow-ai-governance-guide for governance frameworks
• Link to cloud-cost-optimization-guide for infrastructure costs
• Link to agent-architecture-production-ai for scaling AI systems
• Link to what-is-a-fractional-cto for leadership expertise

-->