CIO Priorities

CIOs are bracing for price hikes across spending categories this year as generative AI continues to shape enterprise tech strategies, according to a Gartner report.

Increased technology costs will consume all or most of IT budget growth, according to Gartner. “All major categories are reflecting higher-than-expected prices, prompting CIOs to defer and scale back their true budget expectations,” John-David Lovelock, Gartner distinguished VP analyst, said.

Gartner forecasts spending on software, data center systems and devices to grow by double-digits year over year, due largely to generative AI hardware upgrades. Investments in AI-optimized servers will account for twice the spending on traditional servers, the analyst firm said.

Generative AI is pumping up costs across the board as vendors double down on the technology. The race to embed copilots and agents in software services and invest in GPU compute capacity is adding to enterprise bills, whether CIOs like it or not.

“CIOs were buying things they were interested in last year,” Lovelock told CIO Dive. “This year, they're going to be sold PCs with AI in it, software with AI in it and services with AI in it — they're going to be drowning in AI options.”

Gartner expects worldwide IT spend to grow by 9.8% to $5.6 trillion, an uptick over last year’s 7.7% growth rate. While generative AI remains a key growth driver, strategies and sentiments have changed.

Enterprises focused last year on proof-of-concept work and experienced higher than expected failure rates as they ran into data deficiencies, skills gaps and disappointing ROI.

“From a hype cycle point of view, we're on the wrong side of the peak with all those failures,” Lovelock said. “We now have expectations for the technology coming down. CIOs, CEOs and boards are still talking about generative AI but expectations have come way down.”

For tech services companies, spending patterns reflected a far rosier perception.

Investments in data center hardware and software hit an all-time high in 2024, growing 34% year over year to $282 billion, according to Synergy Research Group. The firm traced more than half of all infrastructure spending back to hyperscale cloud providers.

Gartner sees the trend continuing, with IT services companies and hyperscalers accounting for over 70% of tech investments in 2025.

New year, same problems.

CIOs spent 2024 figuring out how to respond to fast-moving technology demands, grappling with a dizzying spread of solutions from vendors and shifts in global regulations.

As 2025 rolls in, executives remain mired in the same challenges. However, their approach will differ this year as new strategies come into the fold. Leaders will apply FinOps practices beyond the cloud, turning to their vendors to leverage agentic AI and build digital literacy to solve staffing shortages.

Here are 5 trends that will shape CIO roles this year:

1. FinOps expands its scopes beyond cloud

FinOps was born in the cloud, as IT teamed up with finance to put a lid on rising enterprise cloud bills. The approach worked so well that FinOps is spreading organically across the IT ecosystem to help rein in software spending, bringing cost optimization practices to the data center.

AI is also on the FinOps radar, according to Jay Litkey, SVP of cloud and FinOps at Flexera and FinOps Foundation governing board member. 

“When FinOps originated, it had the word ‘cloud’ in front of it,” Litkey said. “Cloud doesn’t preface FinOps so much anymore, even when it’s used by the FinOps Foundation.”

Last year, IT and software asset managers forged alliances with cloud-focused FinOps teams to rationalize SaaS procurements and avoid the sting of costly vendor audits. The FinOps Foundation added data center costs, another escalating category of spend, to its scopes in November.

Enterprise software procurements remain ripe for optimization, according to Litkey. Many organizations are overpaying major providers like Microsoft, Oracle and SAP by neglecting licensing portability provisions.

 “There’s a gazillion dollars being overspent right now on software in the cloud that people don't need to be buying, if they just apply their entitlements,” he said. “Those providers have terms that say on-prem licenses can be used in the cloud.”

As enterprises feel the pain of rising AI costs in the absence of clear returns, executives are beginning to worry.  

 “AI is the new cloud — everybody's doing it but nobody knows how much of it they're doing,” Litkey said. “If you asked executives how much they were spending on cloud eight years ago, they couldn’t tell you. The same thing is happening with AI right now.” 

2. Vendors will lead an AI agent proliferation

AI providers have not been shy about where they believe enterprise adoption is headed: toward agentic capabilities and services. 

Rather than responding to prompts and offering advice, AI agents can decide what to do next without needing step-by-step human guidance.  

Last year, Meta, Microsoft, Google, SAP and Salesforce were among a batch of vendors that began shifting messaging toward AI agents and related use cases. In 2025, the list of options for enterprises to choose from is expected to grow. 

“Agents range from simple prompts to fully autonomous,” Hadley Griffin, a product manager in Office AI at Microsoft, said during an annual shareholder meeting in December. “They assist with everything from prioritizing leads, scheduling meetings, fulfilling orders and many others.”

Agent-focused predictions peppered vendor-published trends for 2025. Salesforce says this year will be characterized by multi-agent systems taking center stage, and Google predicts agentic capabilities will simplify complex tasks for employees as the technology becomes more popular in businesses. 

The shift for vendors from copilot to agent also aligns with increasing pressure from enterprise customers to reach ROI on AI initiatives at a faster pace. Many organizations grappled with roadblocks for the past two years, and agents can offer more direct business value compared to the legwork required with general-purpose models. Even so, there are additional hurdles on the horizon.

While CIOs are interested in agents, nearly 9 in 10 admit their existing tech stack needs upgrades before deploying the technology, according to a survey of more than 1,000 tech leaders commissioned by Tray.ai and published in December. 

 

3. CIOs pursue the CEO seat as tech becomes more critical

The CIO role spent much of the previous decade gaining strategic importance. Organizations expect their technology strategies to drive financial outcomes, prompting more CIOs to align their careers toward the chief executive seat. 

Nearly one-quarter of CIOs chose CEO as the top C-suite position they were likely to pursue in the future, according to Info-Tech Research Group. 

"Technology has built a foundational layer under leadership and business," said Brittany Lutes, research director in the firm's CIO practice. "Those skill sets that the senior-most individual in the organization needed have shifted to be technology focused."

The CIO's growing scope of influence is already shifting reporting structures, according to Gartner VP Analyst Daniel Sanchez Reina. The firm has tracked an increase in the number of CIOs reporting into COOs, reflecting a closer tie between technology and how businesses operate.

When executives report to CFOs, their remit tends to focus on productivity or profitability. But roles reporting into COOs can have a broader impact on how businesses operate.

"This is what CEOs want," Sanchez Reina said. "CEOs want CIOs to have a much wider perspective of the enterprise."

4. AI governance headaches will persist

CIOs found it challenging to manage the different aspects of AI initiatives in 2024, and analysts expect AI plans to become more complex this year as the technology and regulatory landscape evolves.  

“Governance is reappearing as a No. 1 priority because organizations realize that they’re never going to be able to do AI at scale if they don’t have a good and strong governance foundation,” Collibra CEO Felix Van de Maele told CIO Dive. “Governance has almost become sexy.”

Enterprises are up against a fast-evolving landscape. While governance can provide enterprises with the proper bumper-like guardrails to ensure adoption is responsible and compliant, most leaders have yet to crack the code. Less than half of enterprises are compliant with existing AI laws or actively working toward compliance, according to an Omdia survey.

The lack of expertise in how to govern AI is setting enterprises up to start the year on shaky ground, especially for those looking to pursue agentic AI integration.

“The risks become a lot higher,” Van de Maele said. “The control and the confidence that you need to do these things become a lot higher, and I think that’s just not where organizations today are.”

5. Executives will focus on building digital literacy around AI

Fatigued by vendor promises and looking to quick wins, CIOs will spend more resources and time in 2025 building up in-house skills to harness the potential of technology tools.

Part of the solution will come from training. Digital literacy, particularly around AI tools, will emerge as a key executive priority in 2025, said Seth Robinson, VP of industry research at CompTIA. 

"We've seen in a lot of our studies [in 2024], surveying HR professionals and hiring managers, that these skills are needed across the entire workforce," said Robinson. "Digital literacy and skills include some working knowledge of what AI is and how it works."

The IT trade group forecasts two-thirds of organizations intend to train current employees to close IT skills gaps in cybersecurity, software, data, and other areas of IT, up from 59% in 2024.

To help enterprises surmount skills barriers, vendors have deployed free and low-cost AI training programs.. One joint effort last year set the goal of upskilling 100 million tech workers in the next decade, in a consortium led  by Cisco and backed by Microsoft, Google, IBM, Intel and SAP.

Building internal AI knowledge will require leaders to embed the technology throughout processes, according to Owl Labs CEO Frank Weishaupt.

"You really need to have [AI] at the forefront so you're not just training by putting in an article in a knowledge base for your employees, but you're changing the language of how you communicate," said Weishaupt. "There's so many positives to it that it really needs to be on the top of everybody's mind."

Sponsored content

By Broadcom

If your organization is planning AI initiatives in 2025, there's a critical budget item you can't afford to overlook: edge AI networking infrastructure. While many companies are rushing to implement generative AI solutions, there are two important points to understand: AI traffic exhibits radically different network traffic patterns and most WANS are not prepared for the change. Considering the underlying network requirements will give your projects a better chance of success.

Why traditional networks fall short

The enterprise networks you've relied on for years were designed for a different era. Traditional web applications follow a predictable pattern: small requests going up, large downloads coming down. This 9-to-1 downstream-to-upstream ratio has shaped how we build networks, from corporate WANs to internet service provider offerings.

AI applications completely upend this model. When you're training AI models or using generative AI tools, you're often uploading massive amounts of data before receiving responses. For example, AI-enabled smart glasses that stream audio and video have a typical upstream-to-downstream ratio of 8:1. These applications create bursty, unpredictable traffic flows that are subject to bottlenecks on the upstream side. A single user prompt might trigger connections to dozens of edge sites and require large file uploads, potentially bringing traditional WAN circuits to their knees.

The hidden complexities of AI traffic

The challenges of AI networking go far beyond simple bandwidth concerns. Consider video conferencing: latency is irritating but usually won’t end the call. But AI applications can have their entire processing disrupted by network delays, to the point where they break. This is particularly true in distributed AI architectures where multiple AI agents collaborate across different locations. Adding to the complexity, most AI traffic is encrypted, making traditional inspection and optimization methods ineffective. Without proper tools, you can't distinguish AI traffic from other flows, making it impossible to prioritize effectively.

AI workloads also demand high-bandwidth, symmetrical connectivity that can handle large and unpredictable data transfers. Your network must be able to adapt in real-time to these changing demands, something most traditional networks weren't designed to do.

Security in the AI era

The rise of AI applications also introduces new security challenges. Sensitive information accidentally sent to an AI model may be used for training, potentially exposing proprietary data. Your network infrastructure needs to provide visibility into employees' use of AI applications to prevent shadow AI activities and protect sensitive information.

To address these security concerns, many organizations are turning to Secure Access Service Edge (SASE) solutions, which combine network security functions with SD-WAN capabilities. SASE provides cloud-based security services like data loss prevention to protect sensitive AI training data, cloud access security broker to control shadow AI usage and end-to-end encryption for AI traffic—all while maintaining the performance levels AI applications require.

Building an AI-ready network with SD-WAN

Creating an AI-ready network requires rethinking your infrastructure strategy. While upgrading WAN circuits may be necessary, implementing SD-WAN technology can help optimize your existing infrastructure. And if your organization already implemented SD-WAN years ago, it may be time to rethink your initial investment. No longer simply a solution to connect branch offices, SD-WAN has evolved into an intelligent, AI-enhanced overlay that optimizes application performance, enhances security through SASE integration and delivers superior user experiences over any underlay link.

Modern SD-WAN solutions can bond multiple network links, including fiber, 5G fixed wireless access and satellite, into a single logical connection and automatically route business-critical traffic over the best available paths. VeloCloud SD-WAN goes a step further to create application-specific network slices, regardless of the underlay, to guarantee performance. This means you can often use your existing network investments more effectively while ensuring AI applications get the bandwidth and low latency they require. 

Intelligent traffic management is another crucial component. Modern SD-WAN solutions can identify AI applications and automatically adjust network resources to prioritize the performance of the most important business applications, including the ability to create dedicated channels for AI traffic when needed. These solutions should also include dynamic optimization capabilities that can automatically adapt to changing conditions and network behavior over time, ensuring consistent performance without constant manual intervention.

Budgeting for success

When planning your AI networking budget, consider it an investment in your AI initiatives' success. Include costs for infrastructure modernization, intelligent network management solutions and security enhancements. Remember that while the initial investment may seem substantial, the cost of inadequate network infrastructure could be far greater in terms of failed AI initiatives and security breaches.

The bottom line

The success of your AI initiatives depends as much on your network infrastructure as it does on the AI technologies themselves. By including AI networking requirements in your project planning and budgeting from the start, you can avoid performance bottlenecks and security risks that could derail your AI initiatives.

Remember: AI networking isn't just another infrastructure upgrade—it's a fundamental shift in how your organization's network needs to operate. 

Most businesses gained ground on their AI goals last year, but returns on AI investments were harder to come by, according to IBM-commissioned research. 

Less than half — 47% — of IT leaders said their AI projects were profitable in 2024, the survey of more than 2,400 IT decision-makers found. One-third said they broke even, and 14% recorded losses.

Most businesses that have yet to reach ROI expect cost savings to occur within three years. Around 44% anticipate turning the corner sooner, within the next two years. 

Generative AI got a bad reputation in 2024 for its elusive returns, which ultimately clouded the technology’s initial hype and spurred enterprise disillusionment. Most organizations pivoted from a wide-eyed adoption sprint to a more deliberate pace, eyeing costs and prioritizing high-value use cases.

Along the way, organizations have also changed the way they gauge return on investment.

Quicker software development, faster innovation and time saved were the most important metrics IT leaders used to calculate AI investments ROI, according to IBM’s report. Far fewer leaders are focused on quantifiable savings.

While technology leaders expect AI projects to bring cost savings in the long term, the price tag associated with the tech stack is already rising. Nearly 4 in 5 U.S. organizations experienced software price hikes last year, according to a July Forrester report. Most technology leaders blame AI and expect the technology to raise software costs again this year. 

Countless analysts rang the alarm on the risk of ballooning AI costs last year. Gartner said enterprises could be miscalculating AI costs by as much as 1,000% as they scale deployments.

Tech executives are working to manage the cost creep by increasing audits, leveraging their purchasing power and turning to open-source alternatives when appropriate, Forrester said, though contract complexities and predicting future needs present additional challenges. 

Governing generative AI initiatives was overwhelming for organizations in 2024. Enthusiasm has waned and skepticism has increased as enterprises forge on. The most common place for Fortune 500 companies to mention AI in SEC filings is in the risk factor section, according to an Arize AI report published in August. 

CIOs enter the new year facing even more pressure to achieve positive ROI on AI initiatives. In response, vendors are marketing applicable use cases tied to new updates and innovations to accelerate adoption.

Enterprises have big plans for AI agents in the year ahead despite looming concerns over security and governance, according to Tray.ai research. 

More than 2 in 5 enterprises plan to build more than 100 AI agent prototypes, the survey of more than 1,000 U.S.-based enterprise tech pros found. Most organizations intend to spend at least $500,000 per year on the initiative.  

Businesses are aiming implementation efforts mostly at the IT service desk, data processing and analytics and code development and testing. Other priorities include cost reduction and increased automation.

While CIOs are still learning how to best navigate the generative AI wave, another emerging technology is creeping up: agentic AI. 

Unlike current tools that require prompts, AI agents can decide next steps autonomously, such as checking calendars for availability without being explicitly asked to do so before coordinating a meeting time. 

Most tech pros anticipate agents will become core to operations over the next 12 months, powering more than 25% of processes by year-end, according to the Tray.ai survey. 

While vendors are stepping in to provide enterprises with a path to implementation, most CIOs and analysts say organizations are not yet prepared for autonomous tools.

Similarly to generative AI, enterprises have high expectations for the technology and readiness gaps loom large. Nearly 9 in 10 IT pros say their organization’s tech stack needs upgrading in order to deploy AI agents, the survey found. 

The introduction of agents will also require CIOs to strengthen AI governance and security, a process already underway at many organizations.

As business leaders head into 2025, most recognize there are challenges ahead in striving to achieve AI goals in the face of fatigue and burnout, according to EY research.

More than half of senior leaders feel like they're failing amid AI’s rapid growth, and a similar proportion reported companywide enthusiasm for AI adoption is on the decline, the survey of 500 leaders found. 

Business leaders said employees struggle with feeling overwhelmed and exhausted by the persistent onslaught of information and developments tied to AI.  

Even with bumps in the road, enterprises are not giving up on AI dreams. 

More than one-third of senior leaders surveyed said their company is planning to invest at least $10 million in AI next year, up from 30% six months ago, EY data shows. But while some positive gains have emerged, CIOs often feel like the deck is stacked against them. 

Leaders blamed infrastructure and data processes for holding their organization back. Nearly two-thirds of business chiefs surveyed by EY highlighted difficulties in keeping workers motivated to embrace the technology, too. 

Most IT vendors have rushed to capitalize on enterprise AI readiness gaps. From tailored tools to customized training, AI providers are working to boost the technology’s adoption while rapidly scaling up capacity to meet demand. 

But what began as a wide-eyed sprint into generative AI has settled into more of a measured march for technology leaders who now view the technology through a more skeptical lens. 

Around 3 in 5 senior business leaders reported growing interest in responsible AI practices over the past year, up from 53% six months ago, according to the EY survey. Around half intend to put a larger focus on the technology’s risks in 2025, and about the same proportion plan to increase required employee training on responsible use.

The start of 2025 signals a renewed challenge for businesses: More people are looking for a job, HR departments are looking to hire more, but the qualifications job seekers say they have aren’t matching company needs, according to LinkedIn’s “Work Change Report.”

The challenge goes both ways, LinkedIn said: Nearly 40% of job seekers are applying to more jobs than ever but hearing back less often. Meanwhile, almost three-quarters (73%) of HR professionals report that less than half of the applications they receive meet all their listed criteria, the platform found.

“The process is more challenging today because we are seeing a ‘skills mismatch’ between professionals and organizations – and that’s because skills and jobs are changing so quickly,” Erin Scruggs, LinkedIn’s head of global talent acquisition, told HR Dive. 

The good news is that “HR leaders are in a position to help steer their organizations in this era of change by leading the charge on upskilling and reskilling,” Scruggs said.

Scruggs’s recommendation: “Many applicants may not meet every single job requirement but can still be an excellent fit for a role, especially if given the opportunity to build their skills once they are in the role,” she said.

Aside from external hiring, “HR leaders can also help their companies build skills from within,” Scruggs added.

This may address another HR problem: Although hiring managers are receiving a “flood of applications,” they’re struggling to keep top performers, according to a September report from Workday.

Employees want to grow in their roles — a key factor motivating them to stay, Workday found. Internal mobility presents a “win-win” situation, as it boosts employee satisfaction and helps companies retain valuable skills and knowledge, a Workday executive stated.

LinkedIn’s research showed that 77% of HR professionals say their organization is prioritizing upskilling initiatives this year in areas such as AI, sustainability and soft skills, Scruggs pointed out.

Global hiring for AI talent has surged more than 300% over the last eight years, with AI and sustainability jobs — such as AI engineers, environmental officers, energy managers and AI researchers — representing the biggest increase in demand, according to LinkedIn.

The proliferation of AI in the workplace requires new skills for different roles, the leader of The Conference Board’s U.S. Human Capital Center stated in a January report. Organizations that don’t prioritize upskilling and reskilling their workers could be caught in a perilous position, a Conference Board exec warned.

If organizations do face AI upskilling obstacles, it might be because of speed. Nearly half of the 1,200 employees surveyed by TalentLMS said they believe AI is advancing faster than their company’s ability to train them, the company’s Jan. 14 report found.

As Gen Z employees take up a larger share of jobs and older workers retire at later ages, onboarding is a challenge for technology chiefs. Generational differences can result in varying levels of comfort with technology, with AI often acting as an accelerator.

CIOs are looking to help all workers adapt to new tools, resorting to the use of buddies and coaches and maintaining a diverse set of training modules — both virtual, in person or a combination of the two. 

The goal is to help workers benefit from the strengths of their counterparts, regardless of age. 

Cybersecurity assessment firm Schellman learned some lessons on technology adoption while upgrading its case management system from older software developed more than seven years ago. The company swapped out a legacy “walled garden” list-based system with a more flexible platform-based architecture. The upgrade meant bridging workforce generation gaps.

Younger workers who can seamlessly switch between different technology platforms in their personal lives may be more comfortable with an ecosystem-based approach where multiple software tools connect with one another. They may grow frustrated using an older system they deem too restrictive. 

By contrast, older workers accustomed to having all the information in one place may struggle to use a platform that requires them to navigate across several tools laid out in a non-linear way, said Schellman CIO Christopher Kouzios.

“Younger workers might say ‘Can you add more features?’" said Kouzios. “Folks who are used to the older platform might say ‘Why is there stuff all over the place? I had a routine, and my routine has now changed.'”

Resistance to change

Xavier Rebeuf, chief product and technology officer at Paris-based banking software firm SBS, said the company wants all its developers to use technology to carry out API tests. Adoption has lagged for some workers over the age of 40 who are accustomed to delegating manual testing tasks, he said.

SBS uses two types of technology to conduct tests, including open source tool k6, and a generative AI testing tool, Rebeuf said.

Some older workers “only want only to specify the test and to delegate it to someone else to execute, which is not the pattern we want to use right now,” said Rebeuf. “There's a learning curve for sure to understand how to code that” using k6, according to Rebeuf. Older workers have also voiced trust concerns related to the generative AI testing tool.

Other firms also report that older workers may be more wary of AI and consequently are less likely to adopt it compared to their younger counterparts. For example, software company Pegasystems recently rolled out a contract summarization AI tool.  Experienced employees accustomed to researching contracts without assistance gave the tool a lukewarm reception.

“They're just used to doing things a certain way. Maybe it's doing all the research on their own, summarizing it, and then making decisions,” said Pegasystems CIO David Vidoni. “They're warming up to it, but I think it comes with just not trusting [it].”

Training older workers on new technologies is especially important because fluid learning abilities — which allow people to pick up new skills without prior context — can decline with age, said Stefan Tams, a professor at the department of information technologies at HEC Montréal. 

The advantage older workers have over their younger colleagues, however, is more crystallized intelligence, or skills and knowledge acquired through experience, Tams said. Both should be valued as companies approach learning and technology adoption strategies for employees, he said.

Reverse mentoring

Companies have devised a few learning approaches to encourage tech adoption across age categories. One way is through a "buddy" system, where older workers are paired with younger ones. 

This approach lets older workers benefit from younger colleagues’ fluency with technical tools, while younger workers can gain from older colleagues’ experience as they seek to gain more value from workplace tools, said Kouzios.

Schellman offers technology training modules that combine in-person and virtual instruction, given that older workers tend to prefer traditional classroom instruction while their younger counterparts may be more willing to use virtual self-service tools. 

In addition, the company pairs onboarded employees with workers familiar with important technologies, said Kouzios.

“The gap bridge is Gen X,” he said, noting that many buddy volunteers are either Gen X or are older millennials — workers who may be equally comfortable working with new and legacy systems.

Similarly, SBS has introduced a system where younger worker technology “champions” assist older colleagues, particularly on increasing adoption of AI-based testing. Rebeuf said the buddy system is helping older workers trust the technology and driving up adoption.

“[The] automatic test case using AI is increasing," he said. "Little by little, people start to trust the tool. We all know that perhaps people feel in danger regarding generative AI, so it gives the confidence that, at the end, what we are looking for is to accelerate our development.”

Employers should increasingly lean on “reverse mentoring” — instances where younger workers help their older colleagues — to help workforces adjust to technological change, argues Christina Janzer, Slack’s SVP of research and analytics, who also oversees an internal research unit called Workforce Lab. 

In September, the Workforce Lab published a study of 5,000 full-time desk workers and uncovered distinct AI personas employers need to understand as they implement AI in the workplace. Among the personas, four out of five maximalists — those who frequently use AI and encourage others to use it — were under age 44. 

Meanwhile, the polar opposite of that persona, called the rebel — most of whom don’t use AI at work and feel it’s a threat to society — skewed older, with more than half aged 45 or older. 

“There's an opportunity and an expectation that Gen Z is going to help the rest of the workforce," said Janzer. "Let’s give Gen Z the opportunity to reverse mentor and help the older generation."