The bill came due
For years, corporate technology leaders ran internal hackathons, distributed AI coding tools to every desk, and championed adoption as a competitive imperative. In 2026, the invoice arrived — and it is larger than the business case that was supposed to justify it.
Global AI spending is projected to total $2.5 trillion this year, a 44% increase from prior-year levels, according to data cited by Fortune. Some companies have reported that their 2026 AI budgets have blown past what they expected at the beginning of the year without producing any corresponding value to the business.
'2026 is the year of everyone finding out that AI is actually really hard,' said Will Sommer, a quantitative modeling and economic forecasting expert at research firm Gartner. 'It's not a free lunch. It requires a lot of thought and effort to get right.'
Sommer warned that companies can easily spend thousands of dollars per head on AI tools whose output is essentially junk and does not bolster productivity. In June, Gartner issued a report warning that AI coding costs would overtake the average developer's salary by 2028, driven by rising token consumption and a shift to consumption-based fees.
Caps, guardrails and tighter defaults
At Samsara, CIO Stephen Franchetti authorized Anthropic's Claude, Google's Gemini, OpenAI's ChatGPT, and the AI coding agent Cursor — then built an internal system to monitor all AI expenses on a daily basis. More recently, Samsara capped usage for some non-technical employees while giving research and development teams more room to experiment.
At Docusign, CTO Sagnik Nandy found that AI code agents were pulling the company's entire code base for context before executing a task. After adjusting the default setting so agents pull only the context relevant to a narrow task, Nandy said token usage dropped by almost 50%.
At Yum Brands — operator of KFC and Taco Bell — Chief Digital and Technology Officer Jim Dausch contended that as many as 95% of tasks assigned to AI tools can be handled by more basic, less expensive models. Yum has responded with additional training and by pushing business leaders to manage digital spending the way they budget for headcount.
'We're trying to kind of democratize where the costs live and how they're managed, so it isn't just an IT line item,' Dausch said.
At Cigna Group, Chief Data, Digital, and AI Officer Katya Andresen authorized more than 70 different AI models for internal use, steering the workforce toward small language models or earlier, cheaper versions for tasks that do not require heavy reasoning. 'The way you really run up costs is you use the most expensive models with no guardrails around them,' Andresen said.
The market has already voted
AI hyperscalers are hearing the complaints. They have responded by rolling out cheaper models and cutting prices — a market signal that enterprise buyers are pushing back on unchecked consumption.
The episode is a textbook lesson in what happens when procurement discipline is suspended in the name of innovation theater. Free enterprise does not punish ambition; it punishes ambition without accountability. The companies now rationing access and demanding measurable returns are not retreating from AI — they are applying the same capital discipline that separates durable competitive advantage from expensive noise. The ones that get the cost structure right will own the productivity gains. The ones that do not will have funded a very elaborate experiment on the shareholder's dime.
