IT leaders’ top 5 barriers to AI success

Artificial intelligence — and generative AI in particular — is fast proving to be a useful solution for increasing productivity across the enterprise but several common barriers to success remain.

The sooner IT leaders can identify and overcome these issues, the faster they will enable their organizations to get more value from AI-based systems.

Here are some of the more challenging barriers that enterprises need to knock down and how IT leaders can go about doing so.

Poor data quality

An IDC survey of 2,920 global IT and business decision-makers, sponsored by hardware vendor Lenovo, showed that data quality issues are the No. 1 inhibitor causing AI projects to fall short of expectations.

To combat this, one-third of the survey respondents say their organizations plan to prioritize improving their data management capabilities.

Ally Financial’s Sathish Muthukrishnan, for one, is tackling this issue by breaking down silos and emphasizing data governance.

“AI relies heavily on the data it processes, so we carefully manage the challenges and risks concerning the broader use of data,” says the chief information, data, and digital officer at the digital banking firm.

“We consolidated 98% of our data in a centralized cloud-native database, which enables us to harness the power of the data,” Muthukrishnan says. “We’ve implemented processes designed to ensure strong data security, robust customer privacy, and rigorous model risk review before deployment, and continuous monitoring of outcomes.”

Data quality issues are a real concern and an actual barrier to AI adoption, but the problem is much larger than the traditional and typical discussion about data quality in transactional or analytical environments, says John Thompson, senior vice president and principal at AI consulting firm The Hackett Group.

“With gen AI, literally 100% of an organization’s data, documents, videos, policies, procedures, and more are available for active use,” Thompson says. This is a much larger issue than data quality in systems such as enterprise resource planning (ERP) or customer relationship management (CRM), he says.

To address the issue of data quality for gen AI, rather than getting a handle on data quality before bringing in AI, organizations need to load information into gen AI and actively interrogate, query, and prompt to find where the information is accurate and up to date and where it lacks accuracy, relevancy, and clarity, Thompson says.

“Gen AI is the tool to find out where information needs to be improved,” Thompson says. “The right way to execute this process is to bring in gen AI and find out how to fix data quality, not the other way around.”

Lack of in-house expertise

A March 2025 report by American Management Association, a professional development organization, recently surveyed more than 1,100 professionals in North America and found that many employees (57% of respondents) feel that they are not keeping up with AI. Less than half (49%) have received training in AI.

Organizations need the infrastructure in place to educate and train its employees to understand the capabilities and limitations of AI, Ally’s Muthukrishnan says.

“Without the right training, adoption and utilization will not achieve the outcome you’re hoping for,” he adds. “While I believe AI is one of the largest tech transformations of our lifetime, integrating it into day-to-day processes is a huge change management undertaking.”

Ally prepared its employees to use AI responsibly by offering training required for all users, an AI playbook, short courses to increase AI fluency, and educational “AI days” open to the entire enterprise.

“The skills gap is only going to grow,” Hackett Group’s Thompson says. “Now is the time to start. You can start with your team. Have them work on test cases. Have them work on personal projects. Have them work on passion projects. [Taking] time for everyone to take a class is just elongating the process to close the skills gap. Gen AI is accessible to all. You can use that accessibility to provide people with an opportunity to learn by doing.”

Experience is a much better teacher than listening, Thompson adds.

“Set up a gen AI environment and let everyone access it for six months to a year,” he says. “Your employees learn more than they ever would in a class or even multiple classes. I am not saying to not train people; you still need people to take classes. But when your employees are hands on with a gen AI environment, you will see the skills gap closing each day.”

Organizations need to emphasize mixing in-house experts who can pinpoint appropriate AI use cases with external talent who have seen how other organizations are using AI, says Jared Coyle, chief AI officer at enterprise software company SAP Americas.

“The in-house knowledge is critical to make sure you integrate with existing systems and processes, and the external talent better helps you fully leverage newer AI capabilities to keep AI systems running smoothly,” Coyle says.

Finding use cases that can compete for resources

Every enterprise has an established list of priorities, many of which do not involve AI or gen AI. To complete for funding, people, and senior management attention, IT leaders will need to make a compelling case for their AI projects.

“One of our biggest challenges is identifying the right business cases — where AI can drive real, measurable value without adding unnecessary complexity,” says Chandra Venkataramani, CIO at outsourcing company TaskUs.

“It’s easy to get caught up in the momentum of generative AI,” Venkataramani says. “Success comes from resisting that impulse and instead zeroing in on areas where the technology can augment our internal capabilities — such as improving productivity, enhancing decision-making, or reducing friction in key workflows. We’ve found that clarity of purpose — whether it’s reducing cost, increasing speed, or improving user experience — is critical before moving forward with any AI initiative.”

Equally challenging is the issue of competing priorities, Venkataramani says. “As a high-growth, client-centric company, there’s constant pressure on IT resources,” he says. “AI initiatives must compete for budget, talent, and executive attention; that’s when alignment becomes essential.”

The company emphasizes shared ownership of AI initiatives across business functions, ensuring that investments are technically feasible and championed by business leaders who understand the return on investment.

“We’ve adopted a mindset that emphasizes experimentation with guardrails: thoughtful pilots, clear KPIs [key performance indicators], and feedback loops,” Venkataramani says. “This approach has helped us stay agile, avoid over-investment in unproven solutions, and focus on long-term value rather than short-term hype.”

Business cases, use cases, and competing priorities “have been with us since the beginning of our careers,” Hackett Group’s Thompson says. “The great thing about gen AI is that it applies to most problems. I recommend starting with the most pressing problem, the most strategic issue the C-suite is troubled by and start there. I would not spend a great deal of time on business cases or use cases. Dive in. The time to make a difference is now.”

Outdated legacy systems

Many enterprises have launched digital transformation initiatives to enhance efficiency and improve services to customers and employees. Those that haven’t or are lagging in their efforts need to make this a priority, because outdated legacy systems and applications are potential roadblocks to AI success.

Decades-old applications that were designed to retain a limited amount of data because of storage costs at the time are unlikely to integrate easily with AI tools, and in many cases outdated applications are completely blocking AI adoption.

“Success does not come by just attaching the LLM [large language model] du jour and lakehouse technology of your choice together and hoping it all works out,” SAP Americas’ Coyle says. “Many leaders’ passion about AI’s potential is incredible, but focus is really the key. It’s important to avoid launching too many initiatives without the resources to support them.”

Veho, a provider of shipping and logistics services, makes heavy use of AI and machine learning in building and pricing delivery routes and improving the quality of deliveries, says Fred Cook, co-founder and CTO. But to get the most out of AI tools the company had to upgrade its systems.

“Veho’s original core platforms were developed in 2017 and were generally quite brittle,” Cook says. “In late 2023 and throughout 2024, we rebuilt all of our core systems. As they came online, we found that AI applications were much easier to build into our tech, and with that full refactor complete, today we are moving much faster with AI experimentation.”

Veho started with use cases such as a customer support AI agent, driver-partner support AI agent, and client support AI agent. “We have also built some simple AI agents for things like alerting, parsing data, [and quality assurance] of various parcel delivery process steps that would be tedious and manual otherwise,” Cook says.

Another potential hurdle related to infrastructure is the high costs involved. Integrating AI can be any expensive endeavor. Depending on where they are at with IT modernization, organizations might face expenses for systems integration, custom software development, the creation of application programming interfaces, and legacy system upgrades.

Sabotage by employees

A recent survey of 1,600 knowledge workers — 800 C-suite executives and 800 employees — by generative AI services provider Writer 2025 and independent research group Workplace Intelligence finds that 31% of workers had admitted to actively sabotaging their organization’s AI efforts.

Worker sabotage of AI efforts is “a serious issue that can undermine AI initiatives and lead to wasted resources and missed opportunities,” says Orla Daly, CIO at Skillsoft, a provider of education services and technology.

“This sabotage often stems from fear of job displacement, lack of understanding of AI’s benefits, or resistance to change,” Daly says. “To address this, organizations should be taking the time to understand the concerns within their organization, foster a culture of continuous learning, [and generate excitement] around change by engaging employees in AI initiatives.”

Enterprises need to identify AI champions within the organization and share examples of where AI can create a more positive employee experience such as less time spent on administration and more with customers, Daly says.

“Effective leadership is crucial in this process,” Daly says. “Leaders must understand AI and articulate its positive impact on talent and their roles.” This requires a balance of technical, communications, and leadership skills, she says. “When employees see their leaders utilizing AI to augment work with clear use cases and results, they are more likely to embrace AI themselves, turning fear into fascination,” she says.