CIOs cut IT corners to manufacture budget for AI

IT leaders worth their salt know how to make tough decisions, and right now, finding funding for AI projects when budgets aren’t growing is testing the bounds of their executive acumen.

Budget constraints are a fact of life, but with pressure from the C-suite and boards to make AI a priority, IT leaders are feeling extraordinary tension. Often, they must reallocate funds, delay system refreshes, and consolidate vendors and tools while balancing risk and innovation.

“AI spending is moving faster than budgets can realistically keep up with, and most CIOs aren’t ‘finding’ money so much as taking it from somewhere else,” says Kayla Williams, founder and AI and business operations consultant at Twisted Consulting. “The uncomfortable truth is that almost every AI initiative being funded right now is displacing something that was already planned.”

Alex Bakker, distinguished analyst and director at research firm ISG, says the difficult tradeoffs occur when prioritization comes into play. As a result, “organizations that want to grow AI inevitably need to either apply their modest budget growth overwhelmingly into AI, or they have to find an internal budget to reallocate,” he says. 

Reallocations are also time consuming, he adds, and organizations are having to take measures such as decommissioning old apps and paying off technical debt to free up funding for AI, Bakker says.  

Most often, Williams sees organizations putting long-term optimization projects on the back burner in favor of AI initiatives. “Infrastructure cleanups, system refactors, and non-urgent platform upgrades are getting pushed out because they don’t show immediate business impact,” she says. “Those projects matter, but when budgets are tight, they lose out to AI efforts that promise near-term efficiency or headcount relief.”

There’s also a noticeable shift in how corners are being cut, Williams adds.

“Instead of building ideal, future-proof solutions, teams are accepting narrower implementations and more technical debt,” she says. CIOs are greenlighting smaller AI deployments, fewer integrations, and less customization, with the understanding that they’ll have to make modifications later. “It’s not best practice,” she says, “but it’s pragmatic.”

Setting boundaries

Andrew Nassar, CTO at IC Realtime, a manufacturer of video surveillance technology, has been grappling with the AI project requests coming into IT while simultaneously being flooded with near-daily news of new AI tools.

“There’s a lot of wants and needs out there outside of IT and we’re doing our best to combat that,” says Nassar, who budgeted for some tools this year but had nothing to spend on AI tools last year. Now, Nassar is more judicious about what IC Realtime purchases. “We’re taking the stance that the outcome of these tools prove instant results and efficiencies to operations and they’re not crazy experimental right now. We’re setting boundaries.”

That translates into researching and laying out the goals for a particular AI project and what to measure along way. If a project doesn’t pan out, “we’ll icebox it for now,” Nassar says, adding that IT will typically give it one quarter to prove its merit.

One project that got iced was a big initiative proposed at the end of 2025 by IC Realtime’s customer support team to reorganize the organization’s roles. Part of that involved implementing an autonomous sales agent platform with a support chatbot that would answer questions in real-time and point customers to support articles. However, feedback during a pilot was that customers weren’t finding the support articles, “which resulted in more calls,” he says. And, it would have been “multi-hundreds of thousands [of dollars] to stand this thing up.”

With any initiative, you need to understand the tech behind it and what’s involved in running it, Nassar notes. “It’s not just you turn it on and it and it goes.” The platform would have required a team to maintain and program it and continuously configure and tweak it, he says. Another consideration was the fact that the chatbot would be dealing with outbound voice calls, and it wouldn’t be a good look if it didn’t adhere to the tone of the company.

Telling the support team the project was getting backburnered was “well received,” Nassar says. It wasn’t just because of budget constraints, “but also just the complexity and … maybe the risk that [the chatbot] could go somewhat rogue.”

Cutting legacy software and staff, and reallocating funds

Data platform provider Unidata is also having to redirect funds to AI-based data collection and analysis software — and making tough financial decisions has become the new normal.

“We’re making tough budget cuts to legacy software subscriptions and merging redundant tools to reallocate funds to AI, which has forced us to think outside the box about what we can and can’t live without,” says Hanna Parkhots, data collection team lead. “This means something has to give, and for us, it’s cutting back on comfort-zone tools we’ve used for years.”

Company officials opted to reduce the budget for traditional data validation software by 40% and merged three separate project management tools into one, Parkhots says. “This has given us a total budget savings of around $47,000 per year, which we’re now using to fund our AI-based quality control software that can analyze crowdsourced data 73% faster than our old manual process.”

The hardest cut of all, she says, was reducing the budget for traditional data analyst contractors by 30%. “Instead, we’re allocating about $85,000 in funds to AI software to help supplement what’s left of our internal staff.”

Unidata is also “taking shortcuts on training budgets for legacy systems” because it plans to eventually phase them out. Instead, that money is going directly toward AI development and staff upskilling on new tech, she says.

One of the surprising places where cost cuts have been made is in reducing the disaster recovery testing cycle from quarterly to semiannual, saving about $12,000 in contractor and internal labor costs, according to Parkhots.

“The truth hurts: AI isn’t getting a budget increase; it’s taking from everything else,” she says. “We’ve put a hard rule in place that every new AI project must find an equivalent budget cut elsewhere. … This is a zero-sum game that forces us to reevaluate what adds real value versus what we’re just doing out of habit.”

Delaying infrastructure improvements, shelving other projects

Another strategy IT leaders are adopting is to delay non-critical infrastructure improvements. Paul DeMott, CTO of digital marketing agency Helium SEO, says server capacity expansion and network improvements were put on the back burner for 12 to 18 months “because existing infrastructure was adequate.” That freed up about 30% of IT’s annual infrastructure budget for developing AI and paying API costs.

Servers are running closer to capacity limits, he admits, “but the AI tools do make more value than marginal improvements in performance would have.”

Parkhots echoes that, saying Unidata has also put off upgrading non-critical infrastructure for 12 to 18 months. That means no upgrades to network equipment, and no new workstations for administrative staff, she says.

New features not directly related to AI have also been shelved, DeMott says. Helium SEO’s roadmap for this year had “nice-to-have features that would have improved user experience incrementally, but those got shelved to put engineering resources to AI integration,” he says. “Some clients have asked about those delayed features, but when presented with what the AI tools could do, it was a positive reaction.”

Changing traditional resourcing models

Taison Kearney,CISO and data protection officer for professional services platform provider Kantata, has been “pushing the question” of whether AI can change traditional resourcing models. Specifically, he wants to see whether the technology can enable more junior, lower-cost roles to successfully perform work that previously required more senior expertise.

“In some scenarios, that shift meaningfully changes the cost equation and helps offset rising AI investment,” he explains.

Company officials have also formed an internal AI council that encourages ideas from across the organization to ensure it identifies real, practical opportunities. “Each idea is evaluated against consistent criteria, including tooling requirements, total investment, estimated ROI, business case, and the amount of internal development or change management required,” Kearney says.

Based on those inputs, “the focus has been on making conscious business decisions to invest where AI can drive the greatest efficiency gains and measurable ROI, rather than spreading limited budget across too many experimental efforts.”

In several instances, there have been opportunities to solve use cases by developing internal AI capabilities and leveraging AI platforms already in use and pairing them with internal development, rather than adding new tools or incremental vendor spend, Kearney says.

Overall, Kantata’s approach hasn’t been about simply “finding more budget,” he adds, “but about reallocating investment toward initiatives where AI clearly improves productivity, scalability, and cost efficiency, while deprioritizing lower-impact efforts.” 

Practicing ‘budget discipline’ amid the rapid adoption of AI

Amit Basu, vice president, CIO, and CISO of maritime energy transportation company International Seaways, sees the issue of budget sacrificing differently. In many cases, the pressure is not that CIOs are struggling to free up funding for AI, he says, but that senior management and boards are pushing for rapid AI adoption without “corresponding budget discipline, and often, without sufficient focus on governance, security, and risk.’”

The same rigor is not being applied to AI investments that CIOs apply to other major enterprise initiatives, Basu believes. While CIOs are expected to innovate and experiment with AI initiatives, they don’t operate in stable or predictable environments, he says. Yet the metrics used to evaluate success assume operational certainty.

Most existing KPIs measure output and delivery, as opposed to learning velocity, model maturity, and risk discovery, which Basu says are often more valuable than short-term delivery speed. “Without adapting budget discipline to recognize and reward learning, organizations risk slowing real progress while appearing to move fast.”

Consequently, pilots that generate meaningful insights or reduce future risk, but don’t deliver immediate ROI, are often labeled as failures, Basu says.

That makes the CIO’s challenge “less about sacrificing existing programs and more about being asked to move faster than the organization is ready to do responsibly,” Basu says. “This creates a different tension for CIOs and CISOs: balancing speed with control, innovation with resilience, and executive expectations with regulatory and operational realities. In some cases, the greater risk is not which projects are put on hold, but whether AI is being introduced without adequate guardrails and long-term sustainability.”

IC Realtime’s Nassar agrees, saying that, even as a digital video surveillance company that is comfortable using technology, the risk with AI factors into his decision-making, and AI governance will be front and center this year.

“We’ve hurt ourselves a lot through the years with projects” being on the bleeding edge, Nassar explains, which is why “we have been a little hesitant to go crazy experimental on any of these tools or services.” Instead, they “take it easy,” and look at operational cost reductions and efficiency improvements.

Basu says that rather than diverting budget away from existing initiatives to fund AI, many AI projects require IT to first prioritize investments in infrastructure, data platforms, and security to make those initiatives viable.

In International Seaways’ case, this has occasionally meant pausing or slowing AI use cases, “not because of lack of ambition or funding, but because proceeding without the right foundations would introduce unacceptable operational or cyber risk,” he notes. “In that sense, AI has acted as a forcing function, helping the organization make long-needed investments that ultimately strengthen the broader technology and risk posture, not just the AI program itself.”

Vendor consolidation and tool stack reduction

The advent of AI has also meant organizations are having to consolidate vendors and renegotiate contracts, says Twisted Consulting’s Williams.

“Leaders are aggressively trimming overlapping tools, reducing license counts, or delaying renewals to carve out room for AI platforms or services,” she says. “In some cases, AI spend is justified by positioning it as a replacement for manual work that was previously ‘absorbed’ by already-stretched teams.”

Nassar says he’s been able to leverage his existing tool stack for AI projects, and if need be, increase subscriptions slightly. The strategy is to start small and once a pilot proves itself, scale and spend to add new features.

About 5% to 10% of Nassar’s budget will be devoted to AI this year — and that will probably double or triple next year, he says.

Helium SEO’s DeMott has gone further and “aggressively” reduced the company’s software tool stack as well as consolidated tools, which reduced subscription costs by about 40%. “That savings was [directed] into AI platform spending and more headcounts of engineers,” DeMott says. He has also renegotiated contracts with current vendors to get better rates.

Nassar says the fast pace of AI is creating a curveball. “We don’t know what tools will come next week. That’s the scary part for me,” he says. “You’ve got a new coworker from Claude Anthropic that got launched, and it’s put a lot of pressure on us to make sure that we understand these systems.”

But even as some IT leaders are being methodical when it comes to AI funding decisions, Nassar doesn’t ignore the fact that AI is ramping up. “I don’t think there’s a choice anymore,” he says. “This is the biggest capital expenditure the human race has ever seen.”

The budget decisions CIOs are making aren’t being done lightly, says Williams. “Most CIOs know they’re trading short-term stability for long-term capability. But there’s a growing consensus that delaying AI entirely is a bigger risk than postponing other initiatives,” she says. “Falling behind on AI now creates a gap that’s far more expensive and disruptive to close later.”