For CIOs, the useful shift is from procurement thinking to portfolio discipline. Foundational data cleanup, governance and workflow redesign should be treated more like platform investment. Usage-based model costs need FinOps-style controls, with chargeback or at least visibility by team, use case and business outcome. SaaS price uplifts for embedded AI belong in vendor-management and renewal strategy, not just in innovation budgets. Mixing these categories obscures where value is being created and where spend is merely expanding by default.
The article also points to a governance issue that deserves more attention: unmanaged adoption can create a false sense of progress. Broad access may increase token consumption faster than business value, especially when employees use AI for low-impact tasks or duplicate capabilities already embedded in existing platforms. The management question is not how to cap usage indiscriminately, but which use cases merit scale, central support and architectural integration.
A practical next step is to define an AI cost taxonomy before the next planning cycle: foundation, experimentation, scaled use cases, end-user productivity and vendor pass-through costs. That gives CIOs and CFOs a common language for trade-offs and makes it easier to decide where AI is strategic investment, where it is controllable operating expense and where it is simply vendor margin inflation.
Spending on AI isn’t optional for enterprises. But throwing money at the technology is not the key to unlocking its potential as a driver of efficiency and revenue growth. Budgeting for AI that delivers value is proving complicated as CIOs wrestle with costs they don’t always see coming.
KPMG’s Q2 2026 Global AI Pulse found that just 35% of organizations have full visibility into their AI operating costs. Organizations with full cost visibility were five times more likely to report established ROI than those without it.
Why AI costs are hard to predict
Enterprises do not have an AI budgeting blueprint to follow. Leadership teams are figuring it out along the way, which means surprises and mistakes are inevitable.
“It’s really easy to get surprises that pop up along the way. And it’s very, very hard to put structure and predictability around scaling AI capabilities in a large organization,” Paul Blowers, CIO at Plante Moran, an audit, tax, consulting and wealth management firm, told InformationWeek.
The cost of AI extends well beyond buying a platform or tool. AI needs the right foundation to deliver value, and getting that foundation in place requires investment.
“You need the data layer, you need the context layer, you need the translation layer. And then you could have an AI layer and an activation layer,” said Christine Park, chief transformation officer at Branch, a mobile measurement and deep linking platform. Security and governance, which are essential, add to those foundational requirements.
At Branch, getting that foundation in place included costs associated with cleaning up the databases and knowledge bases, Park said. Additionally, she hired a couple of AI automation engineers to help with workflows.
AI transformation costs go beyond the tools
Budgeting for AI also means accounting for the cost of changing how people work and how the work itself gets done.
“When you budget just for the tool, I think you get into trouble because you have to also budget for the transformation,” Park said. “The tool is access, but transformation is how you reimagine your work.”
Reimagining how teams work takes time and money. Park described the work as “a gnarly project” that includes implementation, configuration and workflow changes. “AI does not solve everything. AI is a tool,” she said.
AI consumption costs are hard to predict
Consumption remains a thorny part of the AI budgeting puzzle. Token prices have fallen, but usage has soared.
“Every time a new model is released, or a new use case is released into production, they’re consuming these tokens at a higher and higher rate,” Blowers said. “As CIOs, we’re all getting our PhD in token modeling and negotiation, but the variables are far from fixed right now.”
He said he expects token pricing models to continue to evolve, with vendors taking different approaches.
AI is driving up costs
As enterprises grapple with AI costs internally, they also have to consider how their vendors’ AI investments affect their costs. SaaS vendors are figuring out how to build AI into their products and monetize it, which is driving up their prices.
“Most leading SaaS vendors are raising prices substantially to fund their AI feature roadmaps,” Blowers said. “And it would be safe to say that many, if not most, CIOs would tell you that these prices are increasing ahead of mature AI features being delivered or ahead of our ability to predict what consumption would look like inside of those SaaS tools and platforms.”
How CIOs are budgeting for AI
With AI costs so difficult to predict, CIOs are beginning to rethink how they budget for the technology. For Park, that means treating AI spending less like a fixed line item and more like a variable cost.
“It’s a consumption model. It’s not a fixed cost like SaaS,” Park said.
For companies that did not put in the upfront work to create a solid foundation for AI tools and workflows, the budgeting conversation might begin with the costs of starting over. That foundation matters. PwC’s 2026 Global CEO survey of 4,454 chief executives found that organizations with strong foundations — described as “responsible AI frameworks and technology environments that enable enterprise-wide integration” — have a three times greater chance of reporting “meaningful financial returns.”
When budgeting for transformation costs, enterprises need to paint a clear picture of what integrating AI means for business. Park argued that focusing too much on efficiency metrics is a mistake.
“We really looked at it as a key R&D item because I don’t think AI should be an efficiency play,” she said. “I don’t believe AI equals people cost. I think that’s the wrong formula.”
Whether enterprises opt for an efficiency play or an innovation play, they will need to budget for the cost of AI enablement and adoption. That goes beyond simply buying an AI tool or platform and telling people to use it.
“Giving people access is not adoption,” Park said.
Of course, as enterprises get that adoption, usage — and costs — can rise. And AI budgets need to account for that growth in usage. At Plante Moran, Blowers is working with his executive peers to budget for everyday AI usage, including productivity tools.
“We plan to evolve that much like FinOps: driving adoption, empowering people with spend visibility, introducing monthly limits with some behavioral nudges,” Blowers said. “Training in and of itself is a key cost control, not just limiting token consumption.”
Plante Moran is also identifying AI use cases to scale. “This is a little bit easier to cost manage, to budget and to predict, because we’re centralizing, and these are going through our normal business casing and SDLC processes,” Blowers said.
AI embedded into SaaS products is the third cost area that Blowers and his colleagues are tackling. That means talking to their vendors about how AI is being embedded into the tools the company already owns and how that affects prices when renewals roll around.
“What we’re trying to do to manage that is to insist on opt-in terms, build proactive roadmaps with our vendor partners and really challenge them on P&L impact before signing up for unit token pricing,” Blowers said.
At Branch, tool consolidation has become a part of the budgeting approach. “You want to start consolidating the tools to get some savings,” Park said.
The CIO’s role in AI budgeting
The CIO is a central figure in AI adoption and budgeting, but the job requires collaboration with their C-suite peers and boards to agree on what value they expect from AI and what they’re willing to spend to get that value.
That makes the CFO a particularly important partner.
“You need a close partnership with CFOs,” Blowers said. “I think a healthy tension there is useful, quite frankly. A CFO is an advocate for scrutinizing AI budgets.”
The pressure on CIOs extends beyond controlling costs. CEOs and boards also expect their IT leaders to manage AI risk while moving quickly to keep pace with competitors.
Blowers described, “tremendous pushback on the cost side, tremendous pushback on the risk side and tremendous pressure to go faster and do more to lead and make sure we’re not falling behind. ”
“The CIO’s role is to bring all those concerns together into our AI strategy and help create balance in that discussion,” he said.
Enjoyed this article? Sign up for our newsletter to receive regular insights and stay connected.