Spending on AI is not non-obligatory for enterprises. However throwing cash on the know-how will not be the important thing to unlocking its potential as a driver of effectivity and income development. Budgeting for AI that delivers worth is proving difficult as CIOs wrestle with prices they do not all the time see coming.
KPMG’s Q2 2026 International AI Pulse discovered that simply 35% of organizations have full visibility into their AI working prices. Organizations with full value visibility have been 5 instances extra more likely to report established ROI than these with out it.
Why AI prices are onerous to foretell
Enterprises would not have an AI budgeting blueprint to observe. Management groups are figuring it out alongside the way in which, which implies surprises and errors are inevitable.
“It is very easy to get surprises that pop up alongside the way in which. And it is very, very onerous to place construction and predictability round scaling AI capabilities in a big group,” Paul Blowers, CIO at Plante Moran, an audit, tax, consulting and wealth administration agency, informed InformationWeek.
The price of AI extends effectively past shopping for a platform or instrument. AI wants the fitting basis to ship worth, and getting that basis in place requires funding.
“You want the information layer, you want the context layer, you want the interpretation layer. After which you might have an AI layer and an activation layer,” mentioned Christine Park, chief transformation officer at Department, a cell measurement and deep linking platform. Safety and governance, that are important, add to these foundational necessities.
At Department, getting that basis in place included prices related to cleansing up the databases and information bases, Park mentioned. Moreover, she employed a few AI automation engineers to assist with workflows.
AI transformation prices transcend the instruments
Budgeting for AI additionally means accounting for the price of altering how individuals work and the way the work itself will get accomplished.
“Once you finances only for the instrument, I feel you get into bother as a result of you must additionally finances for the transformation,” Park mentioned. “The instrument is entry, however transformation is the way you reimagine your work.”
Reimagining how groups work takes money and time. Park described the work as “a gnarly undertaking” that features implementation, configuration and workflow modifications. “AI doesn’t resolve every part. AI is a instrument,” she mentioned.
AI consumption prices are onerous to foretell
Consumption stays a thorny a part of the AI budgeting puzzle. Token costs have fallen, however utilization has soared.
“Each time a brand new mannequin is launched, or a brand new use case is launched into manufacturing, they’re consuming these tokens at a better and better charge,” Blowers mentioned. “As CIOs, we’re all getting our PhD in token modeling and negotiation, however the variables are removed from fastened proper now.”
He mentioned he expects token pricing fashions to proceed to evolve, with distributors taking totally different approaches.
AI is driving up prices
As enterprises grapple with AI prices internally, in addition they have to contemplate how their distributors’ AI investments have an effect on their prices. SaaS distributors are determining the right way to construct AI into their merchandise and monetize it, which is driving up their costs.
“Most main SaaS distributors are elevating costs considerably to fund their AI characteristic roadmaps,” Blowers mentioned. “And it might be protected to say that many, if not most, CIOs would let you know that these costs are growing forward of mature AI options being delivered or forward of our skill to foretell what consumption would appear like inside these SaaS instruments and platforms.”
How CIOs are budgeting for AI
With AI prices so troublesome to foretell, CIOs are starting to rethink how they finances for the know-how. For Park, which means treating AI spending much less like a hard and fast line merchandise and extra like a variable value.
“It is a consumption mannequin. It is not a hard and fast value like SaaS,” Park mentioned.
For corporations that didn’t put within the upfront work to create a stable basis for AI instruments and workflows, the budgeting dialog would possibly start with the prices of beginning over. That basis issues. PwC’s 2026 International CEO survey of 4,454 chief executives discovered that organizations with robust foundations — described as “accountable AI frameworks and know-how environments that allow enterprise-wide integration” — have a 3 times better probability of reporting “significant monetary returns.”
When budgeting for transformation prices, enterprises want to color a transparent image of what integrating AI means for enterprise. Park argued that focusing an excessive amount of on effectivity metrics is a mistake.
“We actually checked out it as a key R&D merchandise as a result of I do not suppose AI needs to be an effectivity play,” she mentioned. “I do not imagine AI equals individuals value. I feel that is the mistaken system.”
Whether or not enterprises go for an effectivity play or an innovation play, they might want to finances for the price of AI enablement and adoption. That goes past merely shopping for an AI instrument or platform and telling individuals to make use of it.
“Giving individuals entry will not be adoption,” Park mentioned.
In fact, as enterprises get that adoption, utilization — and prices — can rise. And AI budgets have to account for that development in utilization. At Plante Moran, Blowers is working along with his govt friends to finances for on a regular basis AI utilization, together with productiveness instruments.
“We plan to evolve that very similar to FinOps: driving adoption, empowering individuals with spend visibility, introducing month-to-month limits with some behavioral nudges,” Blowers mentioned. “Coaching in and of itself is a key value management, not simply limiting token consumption.”
Plante Moran can also be figuring out AI use instances to scale. “This can be a little bit simpler to value handle, to finances and to foretell, as a result of we’re centralizing, and these are going by way of our regular enterprise casing and SDLC processes,” Blowers mentioned.
AI embedded into SaaS merchandise is the third value space that Blowers and his colleagues are tackling. Which means speaking to their distributors about how AI is being embedded into the instruments the corporate already owns and the way that impacts costs when renewals roll round.
“What we’re making an attempt to do to handle that’s to insist on opt-in phrases, construct proactive roadmaps with our vendor companions and actually problem them on P&L influence earlier than signing up for unit token pricing,” Blowers mentioned.
At Department, instrument consolidation has develop into part of the budgeting method. “You need to begin consolidating the instruments to get some financial savings,” Park mentioned.
The CIO’s position in AI budgeting
The CIO is a central determine in AI adoption and budgeting, however the job requires collaboration with their C-suite friends and boards to agree on what worth they count on from AI and what they’re keen to spend to get that worth.
That makes the CFO a very vital associate.
“You want an in depth partnership with CFOs,” Blowers mentioned. “I feel a wholesome rigidity there may be helpful, fairly frankly. A CFO is an advocate for scrutinizing AI budgets.”
The strain on CIOs extends past controlling prices. CEOs and boards additionally count on their IT leaders to handle AI threat whereas shifting rapidly to maintain tempo with opponents.
Blowers described, “large pushback on the fee facet, large pushback on the danger facet and large strain to go sooner and do extra to steer and ensure we’re not falling behind. “
“The CIO’s position is to deliver all these issues collectively into our AI technique and assist create stability in that dialogue,” he mentioned.
