As worldwide IT spending is forecast by Gartner to hit $6.15 trillion in 2026, we have started to see a shift in focus from "What can AI do for us?" to "How much is AI costing us?" 

Deloitte warns that AI is now the fastest-growing IT expense, consuming up to half of the IT budget at some firms. When employees use five, ten or even more AI tools -- often without the support of IT and with no centralized oversight -- costs skyrocket, in not just dollars but also reputation, control and hidden risks. 

Untracked spend and true cost of AI

To reduce costs, teams often turn to open source models, and companies limit who has access. I hear daily about employees turning to unsanctioned AI tools to do their jobs more efficiently, and risk grows unchecked. Anyone can talk to ChatGPT instantly through Siri on an iPhone. If an employee does that with sensitive company data, however, who is in control? Not you. 

Related:Intuit's chief AI officer on the SaaSpocalypse and disciplined AI

Shadow AI is a growing risk, and poor access and data controls are leading to massive breaches. From malicious code injected into open source large language models like the supply chain LiteLLM compromise, to viral AI tools like Moltbook that leaked thousands of API keys to hackers. IBM reports that the global average cost of a data breach is $4.4 million. AI adoption is outpacing AI governance. 

McKinsey has found that AI usage is up, with 88% of companies now using the technology. If you aren't using AI yet, then you should take heed of philosopher Meister Eckhart's warning that "The price of inaction is far greater than the cost of making a mistake."

Costs to focus on

Per-user pricing is the status quo that the big guys want you to maintain. Their app, their interface and your data. One airline IT executive said, "We spend about $3,500 a month on [about] 60 employees, and we're always hitting the limits and have to wait." Hitting the usage limits is like eating a hot dog and having to stop halfway through, only to get a notice that you have to pay more. 

Speaking from my experience, it kills productivity. You finally get in the groove of something good, and you get the dreaded: "You've reached your plan's limits. Your plan resets in 3 hours." While power users like me hit limits, lighter users barely touch it, yet we cost the same. We're losing productivity and cash. 

I spoke with a leading global IT services and digital solutions provider with more than 3,500 employees in 36 countries. We talked about the increasing cost of AI, and what their customers were seeing and experiencing. They found most tasks being done by AI could be performed by lower-cost models, but access to them in addition to the premium models meant administration overload.

Related:Time for an AI exit strategy: How CIOs are cutting AI waste

Direct implementation of premium models is great for manufacturing and logistics applications where the machines do the deciding and interacting with the data, but we've seen a spike in costs there as well. I spoke with a popular yogurt company working with one of the big providers on GenAI automation in its manufacturing and logistics processes. He said his company had to spend millions just to get the data set up correctly for model ingestion. 

What happens when a competing provider has a better-performing or lower-cost solution? You're locked in, having blown your budget on data management. Hopefully, there's some reuse there.

The issue isn't with AI itself 

Big vendors want you to lock into their models, their pricing and their data silos. Everyone is pitching their own AI as the best at this and that, but DeepSeek proved you don't need billions to compete. There are some basic steps you can take today to avoid the worst issues and position your company for the best that these tools can bring. 

Step 1. Tackle shadow AI head-on.

This is a cost you are already paying for in risk and inefficiency. Talk to your employees, perform a full audit and collect feedback on what tools they are using, both on and off the books. The goal here isn't to eliminate or restrict employees in what they are using, but to identify and provide sanctioned options that eliminate the need to seek external tools. 

Related:CIOs need control before AI gains accountability

Step 2. Build for flexibility, not lock-in

Can you switch models? Do you keep control of your data? Can you integrate multiple providers or tools? If the answer is no, move on. We are seeing the rise of many third-party providers that are integrating multiple options into a single tool or interface. 

Step 3. Prioritize governance over gatekeeping

The tech industry moves quickly. Your goal is to stay flexible and secure, and shift from "How do we adopt AI?" to "How do we manage AI?" Knowing how your team uses AI could also slash SaaS bills by 20%-40 % through AI consolidation. The companies reaping the most benefits from AI aren't the ones that are spending the most. They are the ones spending smartly. Up-front governance will reap long-term benefits. 

The AI hangover is real, but preventable.

The AI spend is hitting companies across every sector. The rise of data centers to support AI hits everyone in electricity and transportation costs, increasing 7% through December. Utilizing AI smartly will help all of us ease those burdens. It starts with honest conversations about what you are paying for, the benefits, how and why shadow AI thrives, and what pricing models reward productivity instead of penalizing it. 

About the Author