Enterprise AI adoption has hit an inflection point — but not the one vendors promised. While 88-90% of organizations now use AI in at least one business function, only 5-6% report transformative returns from their formal AI investments. The remaining 80-95% of AI projects are failing to deliver their promised ROI or have zero measurable impact on profit and loss.
The gap between adoption and value is forcing a reckoning across boardrooms. CFOs want numbers. CIOs need proof. And the shift to token-based billing in Q1 2026 just made the AI spend impossible to hide.
The Token Billing Reckoning
For most of the generative AI era, enterprise pricing was subsidized and opaque. Flat-fee subscriptions absorbed unlimited token burn. The actual cost of any given task remained invisible to finance teams.
That changed when Anthropic and OpenAI quietly moved enterprise customers to token-based billing. The shift turned a diffuse budget line into a measurable, per-task cost. What it revealed is making investors uncomfortable.
Uber is the most public case. The company burned through its entire 2026 AI coding tools budget by April after rolling out AI tools at near-total scale across its engineering organization. COO Andrew Macdonald acknowledged at a May 25 conference that despite 95% of engineers using AI monthly, he could not draw a line between that token spend and meaningful consumer-facing product improvements.
"That link is not there yet," Macdonald said.
Microsoft faced a similar reality check. Claude Code bills ran $500 to $2,000 per engineer monthly. The company began canceling direct Claude Code licenses and routing engineers back to GitHub Copilot.
One enterprise customer accidentally spent $500 million in a single month on Anthropic's models after failing to set spend limits. The same week, Anthropic closed a $65 billion Series H at a $965 billion valuation.
The gap between those two numbers tells the real story of enterprise AI in 2026.
The ROI Problem Has Two Layers
First layer: output quality. LLMs hallucinate, loop, and fail in ways that are difficult to predict. Every failed run costs tokens regardless of outcome. There is no standard unit for measuring the cost of an AI task because the same task can consume wildly different token counts depending on the prompt, the model version, the context window, and whether the agent makes wrong turns.
Second layer: pricing opacity. Token-based billing made the spend visible without making it legible. CFOs can now see the line item. They still can't explain what they're buying or why the cost varies 10x from month to month.
Gartner projects AI agent software spending will hit $207 billion in 2026, up 139% from 2025. That trajectory assumes enterprises continue to expand AI spend. The Uber signal — and the pattern of companies quietly pulling back token consumption — suggests the trajectory is under pressure.
GitHub Copilot's June 2026 move to token-based billing provided the clearest retail-level evidence yet. Users on the promotional tier reported burning 30 to 60% of monthly credits in a handful of prompts. One user said Copilot went from their favorite subscription to their most stressful overnight.
These are developers — the early adopters cohort with the highest AI literacy and the strongest motivation to make AI tools work. If the cost-value calculation is breaking down for them, the enterprise rollout projections are built on shakier ground than the valuation multiples suggest.
The Shadow AI Paradox
A key factor contributing to the ROI gap is the prevalence of "shadow AI." Over 90% of organizations have employees regularly using personal AI tools like ChatGPT, Claude, Gemini, and Copilot for work — often without official sanction.
These unsanctioned uses are, in many cases, delivering the productivity gains companies seek. But these benefits remain unmeasured, ungoverned, and unscalable at an enterprise level.
The irony: employees are getting value from $20/month consumer subscriptions while enterprises spend millions on formal AI initiatives that deliver zero measurable ROI.
What's Actually Working (And Why)
According to CIO.com's 25th annual State of the CIO survey, which canvassed 662 IT leaders and 249 line of business users, less than a fifth (19%) say AI initiatives have met or exceeded business goals. And 18% admit fewer than a third of AI use cases are meeting defined expectations.
The barriers are clear:
- 32% cite ill-defined ROI metrics
- 31% cite murky corporate AI strategy
- 40% cite lack of in-house expertise
Organizations that ARE achieving stronger returns from AI are combining generative AI with predictive analytics, machine learning, optimization, and agentic systems. Critical success factors include:
- Strong leadership support — not just budget, but C-suite accountability
- Investment in training — profession-specific tooling, not generic chatbots
- Clear governance frameworks — who owns outcomes, who measures value
- Stage-gated funding — tied to outcome milestones, not deliverable milestones
"We don't fund 'build a model,' we fund 'reduce returns by 8% on this category' with checkpoints at 90, 180, and 270 days," explains Thomas Prommer, a longtime CTO and CAIO. "Projects that miss two checkpoints get killed. We kill roughly a third of what we start and that's healthy."
Several enterprise AI use cases are consistently demonstrating measurable ROI in 2026:
Agentic AI: Klarna replaced 853 full-time equivalent employees with a single customer service AI agent. JPMorgan runs over 450 agentic AI applications daily, yielding significant savings and productivity.
Predictive Demand Forecasting: AI-driven models continuously learn from real-time signals for more accurate predictions.
Intelligent Document Processing (IDP): Automating the extraction and processing of information from documents.
AI Productivity Assistants: Helping employees with tasks like information retrieval, document summarization, and drafting responses.
AI Customer Service Agents and Personalization Engines: Enhancing customer experience and driving conversions.
AI Fraud Detection and Cybersecurity: Providing real-time risk scoring and threat detection.
What CIOs Need to Do Now
The shift from AI experimentation to AI accountability is happening fast. Cross-functional steering committees and specialized task forces are emerging as critical building blocks. Eighty-three percent of IT leaders surveyed confirmed their organizations either have such structures in place or are planning to implement them within the year.
Formal processes for approving AI projects are far less evolved. Slightly more than half (53%) of respondents have established some type of official approval process, with 28% planning to activate something within the next 12 months.
KPIs are also not well defined in most enterprises. Less than half (47%) of respondents have established formal metrics, with another 34% planning to do so within the year.
For those measuring AI success:
- 40% track operational efficiency and process improvement
- 34% track employee productivity
- 30% track cost reduction
- 27% track revenue or growth impact (lowest priority)
First Student, a leading provider of school bus transportation services, stood up a well-defined innovation framework and AI-specific council. CIO Sean McCormack credits these moves to the firm's early success scaling AI initiatives aligned to key business goals.
"We have more discipline around business cases than most companies," says McCormack. "Everything is metrics-driven and dependent on proving value. By the time we put something into production, it's been through a series of proof of concepts, there's been a deep dive on financials, and we are able to move quickly and demonstrate value."
Three years into its AI journey, TIAA has a rich stable of generative AI and agentic AI use cases spanning fraud detection and prevention along with call center companions. The majority (85%) of the financial services firm's workforce uses TIAA Gate, its internal AI platform.
Yet even with all the right structures in place — investment in training, robust governance frameworks, steering committees, an AI center of excellence (CoE), and an enterprise mandate for strategic use of AI as part of everyone's performance goals — ROI remains a challenge.
"What's on paper sometimes doesn't turn into real ROI given the reality of operational costs," notes Sastry Durvasula, TIAA's chief operating, information & digital officer. "Something may prove to be a successful pilot, but you need to understand the full cost of operations — for example, the efficiencies of running tokens or how you're handling traffic or RAG [retrieval augmented generation]."
The Three Recommendations That Work
Based on conversations with CIOs and technical leaders, three recommendations consistently emerge:
1. Establish joint accountability. Name both a technical and business sponsor for each project. Both co-own outcomes. Replace centralized AI CoEs with embedded AI squads that live inside individual business units. Embedded teams force accountability at the point of business impact.
2. Implement stage-gated funding. Tie funding to outcome milestones, not deliverable milestones. "We don't fund 'build a model,' we fund 'reduce returns by 8% on this category' with checkpoints at 90, 180, and 270 days," explains Prommer. "Projects that miss two checkpoints get killed. We kill roughly a third of what we start and that's healthy."
3. Engineer the experience layer. Develop a keen understanding of business workflows and engineer the experience layer for the people tasked with executing AI-enriched workflows. "If someone on the data science team builds a great model that provides insights on improving manufacturing efficiency, but it's so far removed from what the shop floor supervisor does in day-to-day life, it will never be used at scale," says Sriram Krishnasamy, the former chief digital information and transformation officer at FedEx.
The Bottom Line
Token-based billing is the first real price discovery mechanism the AI industry has produced. Flat-fee subscriptions created convenient optics: costs were low, adoption was high, and ROI was a question for later.
Usage-based billing makes ROI the question now.
Anthropic's path to justifying a near-trillion-dollar valuation runs directly through enterprises proving, to their own finance teams, that tokens are worth buying. The companies that can measure that return first will determine whether the current capital stack holds.
The companies that cannot will be the first to renegotiate.
Enterprise AI investment momentum continues to build. But the emphasis for 2026 is squarely on demonstrating tangible business ROI through strategic implementation, effective governance, and a clear understanding of where and how AI truly drives value.
The perpetual pipeline of AI pilots is giving way to a new mandate: prioritize and scale AI solutions with the greatest propensity to deliver business value and impact the bottom line.
88% of companies use AI. 95% see zero ROI. The gap is forcing a reckoning — and the companies that figure it out first will own the next decade of enterprise AI.
Continue Reading
- The AI Cost Crisis: Why Token Billing Is Breaking Enterprise Budgets
- Shadow AI: The Ungoverned Revolution Delivering Real ROI
- How Agentic AI Is Finally Delivering Measurable Enterprise Value
Rajesh Beri writes about Enterprise AI for Technical and Business Leaders at THE DAILY BRIEF. Follow on LinkedIn, Twitter/X, and Facebook.
