Vendors have gotten smarter about how to price AI, one expert says. Here’s how buyers can keep up.
Pm Images/Getty Images
AI spend is “starting to become a CFO’s problem,” Connor Augustyn, managing director and partner for finance transformation at consulting firm West Monroe, told CFO Brew. CFOs, he said, are feeling pressure to show returns on their companies’ investments in AI, but in many cases, productivity gains haven’t arrived yet. Efficiency gains from AI by individual workers are “not measurable in a way that’s truly showing up in the P&L at the end of the day,” he said.
At the same time, investments in the technology are rising. In a recent West Monroe survey report, Augustyn said, “you saw…a really interesting dynamic, where everybody is saying that they want to be able to cut SG&A, they want to be able to preserve the bottom line, they want to be able to make their organizations more efficient, but at the same time, they’re all increasing their technology and IT spend bucket.”
The West Monroe survey of 300 US executives, comprising technology buyers and sellers, found that 74% of buyers spent 11% or more of their IT budgets on AI, compared with 23% of buyers in 2025. About half (48%) said their AI costs stayed within budget, while 30% overspent by 1%–20%. (West Monroe fielded the survey in Q3 2026; buyers’ organizations made at least $100 million in annual revenue and sellers’ companies at least $30 million.)
While only 2% of respondents said their AI spend exceeded 20% of their budget, when AI costs jump, it can be dramatic. “It’s happened several times,” Augustyn said. “I get a panicked phone call on a Monday morning” from a CFO hit with a sudden, hefty AI bill. Perhaps they gave their R&D team free rein to use enterprise AI, Augustyn said, and “somebody let a model run for a week long, and now they’re staring down the barrel of having to fork over $150,000 more than they ever anticipated.”
Clients are struggling, Augustyn said, because AI, unlike traditional SaaS, has variable costs. Now, vendors are “a lot smarter about how to price the AI” than when they raced to the market with AI capabilities, Augustyn said. As they add new AI features to their products, their costs rise and they pass those costs along to consumers, he said. Almost all software buyers (98%) in the West Monroe survey said costs had risen in the past year, and 42% increased their budgets in response.
Keeping a lid on costs. Some buyers, Augustyn said, rushed to implement AI and then found they “led an uncontrollable monster into [their] ecosystem.” They then played catch-up, attempting to “overly control what’s happening with these LLMs” and “throttling people’s usage.”
But there are less reactionary tactics for governing AI spend. A first step is to gain visibility into where the tokens are spent. Augustyn counsels all his clients to track token usage by user when introducing AI.
Staff can also practice “model optimization,” Augustyn said, using the most cost-effective AI model, rather than the most powerful and pricey one, for each task. Tools like LLM routers help direct users to the appropriate AI models, he said, and software vendors have also started to integrate model optimization features into products.
Point of leverage. Renegotiating software contracts can be an opportunity for savings. Nine in 10 software buyers in the West Monroe survey renegotiated a major IT or software contract in the past year. That’s happening in part because M&A transactions, which have picked up, can trigger renegotiations, Augustyn said. The largest driver, though, is vendors adding brand new products that can change contract terms.
When heading into such a negotiation, CFOs should have a sense of what their organization’s consumption might be for a given tool, Augustyn said, or risk over- or under-buying. He also suggests negotiating for “price protections,” such as caps on tokens.