What’s holding back enterprise AI? Shortage of talent, CIOs say

A shortage of expertise has held back AI initiatives at many organizations, with shallow knowledge of the technology plaguing practitioners’ ability to make good on the promise of AI.

According to CIO.com’s 2026 State of the CIO survey, lack of in-house talent was the top challenge IT teams faced in implementing AI strategies during the past 12 months, identified by 40% of respondents.

The shortage is especially acute for roles at the intersection of AI and cybersecurity, says Ha Hoang, CIO at cyber resilience vendor Commvault. Cybersecurity companies need people who can understand data and operations and translate risk insights into business decisions, she says.

Vendors such as Commvault also need engineers and analysts who understand how to secure AI models, protect training data, and detect AI-related threats such as prompt injection and model poisoning, she adds.

“As AI-driven automation reshapes IT and security operations, CIOs and CISOs will need professionals who can interpret, tune, and govern AI systems, not just monitor alerts,” Hoang says. “We’ll need fewer siloed specialists and more AI-fluent generalists who can evolve as technology does.”

Deep expertise needed

Part of the problem is a shortage of people who understand the power of AI and can predict where AI technologies are headed, adds Anand Srinivasan, chief strategy officer of AI-powered enterprise planning platform vendor o9 Solutions.

“The challenge is not simply a shortage of AI experts, but a deeper structural gap between how enterprises are organized and what modern AI enables,” he says. “Most large organizations still operate through functionally siloed, hierarchical decision-making models designed for stability and scale, not speed and adaptability.”

The most critical expertise gap is not just in building AI systems, but also in rethinking how decisions are made and executed across the enterprise, Srinivasan says. AI can enable huge changes in agility and adaptability, but only if enterprise decision-making capabilities allow organizations to convert strategy into action faster and with less risk, he adds.

Srinivasan quotes hockey legend Wayne Gretzky to illustrate the problem: “Skate to where the puck is going, not where it has been.” The AI puck is moving very fast, he notes, and AI expertise is a moving target.

“Skills in traditional ML are being rapidly displaced by needs for generative AI, agentic AI, and AI governance,” he adds. “Workers with AI skills now command significant wage premiums over peers in the same roles without those skills.”

Beyond the challenges with a fast-evolving technology, there’s a problem with shallow AI expertise, adds AJ Sunder, CIO and chief product officer at strategic response management software vendor Responsive. There are plenty of people available who have some AI knowledge, but many lack a deeper understanding of how to deploy it to meet enterprise needs, he suggests.

“There is certainly a shortage of people that can build reliable, safe, production-scale AI systems,” he adds. “This abundance of AI-aware talent, combined with a dearth of people that can translate that into functioning AI applications, creates a massive problem sorting through the noise.​​​​​​​​​​​​​​​​”

It’s been a challenge for Responsive to find workers with that level of expertise, but the company has been fortunate to find some outside talent, Sunder says.

“The types of AI problems we solve require expertise in dealing with content at scale, with all the complexities of messy enterprise data,” he adds. “There aren’t too many people with sufficient experience solving the kind of problems we solve at the scale we do.”

Hands-on training

Responsive has prioritized internal training to build in-house expertise, with internal teams driving educational efforts, Sunder says. The AI-focused company had a bit of a head start because it had already focused on the technology before the current wave.

“We’ve been fortunate to have talented people that quickly recognized the pace of AI and the value of hands-on learning, experimentation, trial and error, and unlearning to learn new things,” he adds. “That meant all of us collectively learning, sharing, and teaching one another.”

The company also builds teams by pairing AI specialists with domain experts rather than putting them in isolated groups, Sunder says. Response has also invested aggressively in AI tools that allow a broader set of engineers contribute to AI-powered features without needing deep ML backgrounds.

“You don’t need everyone to be an AI expert right away,” he says.

Sunder questions the need for more outside AI training programs, saying there may already be too many out there.

“Some structured training to bring most if not all of your team members to a baseline level of knowledge is necessary, and it’s out there,” he says. “Beyond that, unstructured learning, hands-on exercises, building useful solutions beyond ‘hello-world’ tutorials are far more effective than any long-running training programs could do. This is mainly due to how fast things are evolving.”

Commvault is also focused on internal training methods and on reskilling current employees, Hoang says. The company is also exploring partnerships with universities and cybersecurity boot camps.

“The hardest skills to find are those that combine security fundamentals with AI model governance or automation tooling,” she says. “Many practitioners have one side of the equation, but not both.”

Companies also need to be flexible about how they view AI expertise, she says.

“Many organizations still rely on rigid job descriptions that overemphasize years of experience or specific certifications, while candidates have transferable skills but lack the exact title or tool exposure,” Hoang adds. “Forward-looking CIOs are rethinking the hiring funnel by prioritizing capability and a learning mindset over narrow experience.”