Why enterprises struggle to measure AI ROI amid rising costs

The AI gold rush of 2024 left many enterprises counting costs instead of reaping rewards. What started as a sprint to integrate generative AI tools devolved into a marathon of budget reconciliation when companies like Uber discovered their annual AI expenditures could vanish in weeks, not months. This abrupt shift from experimentation to fiscal reckoning exposed a critical gap: most organizations still lack frameworks to quantify AI’s return on investment (ROI) meaningfully.

Venture capitalist Tiffany Luck of New Enterprise Associates (NEA) has observed this struggle firsthand. During a recent industry panel, she highlighted how enterprises are still navigating the aftermath of the "tokenmaxxing" era—a period where AI usage was prioritized over financial prudence. "Companies are realizing that pushing AI to its limits doesn’t automatically translate to business value," Luck stated. "The real challenge lies in aligning AI spend with measurable outcomes."

The hidden costs of AI experimentation

The fallout from unchecked AI adoption reveals a pattern of financial missteps. Uber’s reported burn rate on AI tools—reportedly exhausting an entire year’s budget in months—serves as a cautionary tale. Meanwhile, Meta’s decision to dismantle its internal AI leaderboard signaled a shift from performance metrics to cost efficiency. These moves reflect broader industry trends where organizations are forced to reassess their AI strategies.

Several factors contribute to the ROI conundrum:

• Unclear success metrics: Many companies conflate AI usage with business impact. Spending on tools like Anthropic’s Claude without defined KPIs makes it nearly impossible to gauge value.
• Shadow AI proliferation: Disjointed adoption across departments leads to redundant tooling and inflated budgets. Teams operating in silos often duplicate efforts without central oversight.
• Implementation lag: AI projects frequently stall at the pilot stage, with only 20-30% reaching production, according to McKinsey’s latest AI adoption report.
• Vendor lock-in: Enterprise deals with AI providers often lack transparent pricing models, locking companies into unsustainable cost structures.

NEA’s framework for AI ROI accountability

Luck advocates for a structured approach to AI investment, emphasizing three core principles:

1. Start with business pain points: Identify specific problems where AI can deliver measurable impact. For example, automating customer support to reduce response times or optimizing supply chains to cut operational costs.

1. Implement phased rollouts: Pilot AI solutions in controlled environments before scaling. This limits financial exposure while allowing teams to refine models based on real-world data.

1. Track granular metrics: Move beyond usage counts to track ROI drivers like cost savings, revenue uplift, or process efficiency gains. For instance, measuring the reduction in manual labor hours per AI-automated task.

She also warns against the "build vs. buy" trap, urging companies to assess whether off-the-shelf solutions can meet their needs before investing in custom development. "The goal isn’t to deploy AI for its own sake," Luck noted. "It’s to solve real business challenges at a sustainable cost."

The road ahead: From hype to hard numbers

Industry experts predict 2025 will be the year of AI ROI accountability. Companies that fail to adopt disciplined financial frameworks risk repeating the mistakes of 2024—where enthusiasm outpaced execution. Luck anticipates a surge in demand for AI consultants specializing in cost optimization and value assessment.

For enterprises still grappling with AI budgets, the path forward is clear: align spending with tangible outcomes, prioritize transparency in vendor relationships, and resist the urge to deploy AI without a defined purpose. As the dust settles on the tokenmaxxing era, those who adapt quickly will emerge with not just cutting-edge tools—but a sustainable competitive edge.