Briefing at a Glance

  • According to Gartner, released on Wednesdayreport, the rise in token prices will cause AI coding costs to exceed the average salary of ordinary developers by 2028. This shift will occur as major vendorsreplace subscription models with pay-as-you-go pricing.
  • The report notes that the pricing model change also makes AI costs a highly volatile variable, making it difficult for enterprise technology leaders to accurately forecast and control spending. Vendors often lack transparency in how token consumption is calculated and billed.
  • Nitish Tyagi, Senior Principal Analyst at Gartner, said in a statement that the cost structure change comes as enterprises are still striving to mature their AI projects andmeasure their business impact. "Software engineering leaders are increasingly concerned that token-driven AI spending is becoming harder to justify, with budgets often being exhausted earlier than expected."

Deep Insights

Integrating AI into software workflows has become the norm, leading employees to spend less time writing code and more time onmanaging AI outputs。

As AI becomes more prevalent in enterprises, costs continue to accumulate, especially in engineering departments, Gartner found. Token overspending is related to how software engineering leaders manage usage, with many using ungoverned autonomous agents in their workflows.

"AI coding costs will continue to rise as infrastructure investments and profitability challenges push up model pricing," Tyagi said. "Meanwhile, as more developers adopt AI tools, light users are expected to quickly transition to mainstream users as familiarity and reliance increase, further driving token consumption and overall spending growth."

According to data released by Altimetrik in April, few enterpriseshave clear goals and outcome strategies for their AI projects, but they choose to move forward to avoid missing the AI wave. KPMG, released on Wednesday,reportfound that only slightly more than a quarter of enterprise C-suite executives report having full, real-time visibility into the operational costs of their AI systems.

Rahsaan Shears, AI enterprise transformation lead at KPMG LLP, told CIO Dive via email that it is likely many agents run background tasks for days without leaders knowing or auditing them.

"The CFO can't see it, and the CIO may not see it either," Shears said. "That's the current state of enterprise AI economics: costs accumulate in workflows that are not fully monitored."

Shears said that to prevent AI usage and costs from exceeding budgets, CIOs must consolidate and closely track usage, monitoring consumption across cloud platforms, copilots, agent frameworks, coding tools, business processes, and team-level experiments.

Organizations should adopt a token economy semantic model that links usage to cost and cost to ownership. This approach also maps AI ownership to business value, workloads, behavior patterns, and risk.

Shears noted that while overspending is costly for enterprises, the lack of visibility into the operational costs of AI systems is a greater risk. If vendors reduce capabilities or costs surge, technology leaders will be in a reactive position.

Shears advises CIOs to investigate which agents need rate limiting, which frameworks run unnecessary loops on advanced models, and which workflows are mission-critical.

"Without this visibility, organizations are like managing a fleet of agents without gauges or a triage plan," Shears said.