Enterprise technology leaders are confronting an uncomfortable reckoning over artificial intelligence investments, as companies that embraced aggressive token-maximisation strategies now face mounting pressure to demonstrate measurable returns. Tiffany Luck, venture capitalist at NEA, told TechCrunch AI that enterprises remain fundamentally uncertain about their AI return on investment, signalling a critical inflection point in corporate adoption economics.
The revelation comes as businesses transition from experimental AI deployments to production-scale implementations, where token consumption—the unit of measurement for API calls to large language models—translates directly into operational costs. Unlike traditional software licences with predictable pricing, token-based billing creates variable expenses that scale with usage, making cost control and value justification considerably more complex.
“Enterprises are still figuring out their AI ROI,” Luck stated, highlighting a disconnect between the enthusiasm that drove initial investments and the financial discipline now required to sustain them. This uncertainty manifests across multiple dimensions: difficulty attributing productivity gains to specific AI tools, challenges in measuring qualitative improvements like decision quality, and the absence of standardised frameworks for calculating AI-generated value.
The ROI crisis particularly affects early adopters who pursued token-maximisation strategies—deploying AI across numerous use cases to identify value opportunities. These organisations now face budget scrutiny as finance departments demand concrete justification for expenditures that can reach six or seven figures annually for enterprise-scale deployments. The shift mirrors earlier enterprise technology cycles, from cloud migration to mobile transformation, where initial experimentation gave way to rigorous cost-benefit analysis.
Technology vendors offering AI infrastructure and model access stand to lose if enterprises curtail spending or demand more favourable pricing structures. OpenAI, Anthropic, and Google—whose business models depend on token consumption—may face pressure to introduce more predictable enterprise pricing tiers. Conversely, companies providing AI cost management tools, usage analytics platforms, and ROI measurement frameworks are positioned to benefit as enterprises seek greater financial visibility.
The consulting sector also gains from this uncertainty, as organisations lacking internal expertise turn to external advisors for AI value assessment and optimisation strategies. Firms offering AI audit services and implementation guidance can command premium fees during this transitional period.
Industry analysts suggest the current situation reflects broader maturation in enterprise AI adoption. Gartner’s 2024 AI hype cycle positioned generative AI at the “peak of inflated expectations,” predicting a subsequent trough of disillusionment before productive plateau. The ROI pressure Luck describes aligns with this trajectory, as enterprises move beyond proof-of-concept enthusiasm toward sustainable operational integration.
The challenge extends beyond pure financial metrics. Organisations must account for indirect benefits—employee satisfaction from automated mundane tasks, competitive advantages from faster decision-making, and risk reduction through enhanced analysis—that resist straightforward quantification. This complexity explains why many enterprises struggle to construct compelling business cases despite perceiving genuine value from AI deployments.
Several factors will determine how this ROI crisis resolves. Model providers may introduce consumption-based pricing with caps or hybrid licensing models that offer greater cost predictability. Enterprises themselves might develop more sophisticated measurement frameworks that capture AI’s multifaceted value proposition beyond simple productivity metrics. Additionally, competitive pressure could force organisations to maintain AI investments regardless of immediate ROI clarity, treating them as strategic necessities rather than discretionary expenditures.
The coming quarters will reveal whether enterprises double down on AI spending despite measurement challenges, or whether budget constraints force selective deployment focused on highest-confidence use cases. For technology leaders, the ability to articulate clear value propositions—supported by concrete metrics rather than aspirational claims—will increasingly determine which AI initiatives survive financial scrutiny and which face the axe.
