The End of the AI Subsidy and the New Cost Reality
The era of “all-you-can-eat” AI is ending, and enterprises are feeling the shock in their budgets. Anthropic’s recent move to unbundle enterprise pricing—charging a base seat fee and billing token consumption separately—signals a structural shift in how AI services are monetized. As OpenAI leadership has noted, unlimited AI plans increasingly resemble unrealistic unlimited electricity contracts. Once frontier AI labs reprice, those costs flow through to the HR, productivity and workflow platforms built on top of them, showing up at renewal time as higher enterprise software costs. Early data is sobering: the average enterprise AI budget has reportedly jumped from USD 1.2 million (approx. RM5.5 million) per year in 2024 to USD 7 million (approx. RM32 million) in 2026, while most IT leaders report unexpected overages from usage-based AI pricing. This is no longer just a technology adoption issue; it is a core budget and governance challenge.
Agentic HR Stacks and the Hidden Consumption Explosion
HR platforms are rapidly morphing into networks of AI agents, and with that shift comes a surge in token usage that many organizations are not tracking. Major HCM vendors are building agentic AI directly into their core architectures and integrating with external agents such as productivity copilots and large language models. A single agentic workflow can now trigger 10 to 20 model calls, multiplying consumption across recruiting, payroll, performance management and talent processes. Employees also bring their own AI tools to work, further inflating usage. The result is a growing disconnect: leaders can describe the productivity benefits but often cannot answer a basic question—what is our daily token consumption and who is accountable for it? Without that visibility, CHROs and finance leaders risk underestimating the true AI bill, mispricing contracts and making workforce decisions on incomplete cost data.
Per-Seat Pricing Under Pressure and the Rise of Consumption Models
Traditional per-seat SaaS pricing is being destabilized by AI agents that automate tasks previously performed by humans. When an AI system can draft briefs, triage backlogs or respond to queries autonomously, the value delivered no longer correlates neatly with the number of human users. Vendors face a structural trap: the better their AI, the fewer seats customers need. Investors have already signaled concern, as software stocks recently logged their worst quarter since the 2008 financial crisis after a major AI product announcement. In response, vendors are experimenting with new AI pricing models. Many are adopting hybrid structures that combine per-seat fees with consumption-based pricing, as seen in seats-plus-credits approaches tied to AI usage rather than pure headcount. Yet no major provider has fully embraced usage-only or outcome-based pricing, leaving enterprises in a complex middle ground where they must manage both license counts and variable AI consumption simultaneously.
Why AI Budget Planning Is Now a Workforce Strategy Question
Rising AI costs are directly shaping hiring plans, training investments and workforce design. Agentic AI can deliver impressive unit-level efficiencies—such as large-scale virtual HR support or automated customer interactions—but total spend can still climb when volume grows faster than expected. One telecom’s multi-agent system, for example, achieved significant per-action savings while tripling total token volume, illustrating how aggregate usage can erase apparent unit gains. Meanwhile, highly publicized cases of AI-driven cost savings that represent only a small slice of overall expenses show how easy it is to overstate return on investment. For CHROs and CFOs, AI budget planning now sits alongside headcount planning: decisions about where to deploy AI influence which roles are hired, which skills are retrained and how much human capacity must be retained for quality, governance and exception handling. Ignoring this linkage risks both budget overruns and poorly calibrated workforce strategies.
Practical Steps for CHROs and Finance Leaders
To regain control over escalating AI costs, leaders need a coordinated playbook that links AI pricing models to workforce decisions. First, conduct a cross-functional audit of AI tool adoption, focusing on where agentic workflows exist, how many model calls typical processes generate and where shadow AI usage occurs. Second, establish ownership of daily token consumption metrics across major platforms, integrating these into financial forecasting and vendor management. Third, renegotiate contracts with an eye on consumption-based pricing, favoring transparent meters, guardrails and alerts over opaque “unlimited” bundles. Hybrid AI pricing models can offer flexibility, but only if enterprises set clear thresholds and budget caps. Finally, embed AI budget planning into talent strategy: define which processes will be automated, which roles will be redesigned and which capabilities will be upskilled. Treat AI spend not as an isolated IT line item, but as an integrated lever for productivity, headcount and long-term capability building.