3 promises every CIO should keep in 2025

Nearly all tech surprises last year were related to gen AI, which was so hyped in 2023 that every organization had to try it in one or more projects in 2024. IT departments ran proofs-of-concept (PoCs), but some business leaders outside IT with P&L to manage also ran their own experiments without necessarily informing IT when they did so. Sometimes it’s as easy as subscribing to an AI-driven service. “You expect a certain amount of shadow IT, but there was much more of it last year,” says Krishna Prasad, CIO of technology services business at UST.

The trouble is, when people in the business do their own thing, IT loses control, and protecting against loss of data and intellectual property becomes an even bigger concern. Like many organizations, UST began setting up an AI leadership council to make sure the same controls are applied to all undertakings, and that people don’t go too far off script with pet projects. They also improved their AI governance.

While many PoCs ran in 2024, most were disappointing. One challenge was that AI algorithms are never 100% reliable. If a model hallucinates just 5% of the time, that’s enough for most users to consider it useless. Another challenge was IT may not have had the available skills to support the new technology. They couldn’t hire people from outside either, because they hadn’t anticipated the need early enough to put it in their budgets.

Everything happened very fast. While IT leaders generally expect technology to follow hype cycles described by Gartner, gen AI did so at an accelerated pace. “What really surprised me last year was the hype cycle for AI went into the trough of disillusionment so quickly,” says Sidney Fernandes, CIO of University of Southern Florida (USF).

Nevertheless, there were enough success stories for executives to read about in the press and wonder why their own IT departments couldn’t do the same. Forward thinking CIOs quickly realized the need to educate business leaders on AI, and by doing so, some were able to head off further misunderstandings about how much business value could be derived from the current generation of algorithms.

The business value that was enjoyed wasn’t always in the places people expected. Gen AI is primarily applied to three categories of use cases, says Prasad. The first is for personal and organizational productivity. The second is to change business processes. And the third is to use AI to build new revenue-generating products and services.

While most IT leaders struggled to demonstrate success in the second two types of use cases, by the end of 2024, personal productivity applications were regularly paying off, so much that many of them became part of the standard office toolkit. For example, most people now use AI to take meeting notes. According to Fernandes, these success cases are further evidence of an accelerated hype cycle. “To my surprise, by the end of the year, we may have already reached the plateau of productivity,” he says.

Given everything that was learned in 2024, it would be a shame for CIOs not to commit to three things that could help them prepare for the year ahead.

Educate the business on AI and educate yourself on the business.

To minimize disappointment, technologists need to set the expectations of business leaders. At the same time, they need to evangelize on the value of new technology. “The CIO has to be an evangelist, educator, and realist all at the same time,” says Fernandes. “IT leaders should be under-hypers rather than over-hypers, and promote technology only in the context of business cases.”

According to Ron Guerrier, CTO of Save the Children Foundation, one way of helping business leaders learn what’s really possible is to recommend books to read on AI. “You don’t want to let them get most of their information from Google searches and YouTube videos,” he says.

Likewise, in Prasad’s view, IT leaders should step up their own efforts to understand the business. They should become more proactive about meeting with business leaders, and insist that members of their staff meet regularly with their counterparts in the business. IT leaders need to face the fact they no longer add value simply by being technical. They should learn to talk about projects in terms of three business parameters: reducing risk, reducing cost, and increasing revenue.

According to Leon Roberge, CIO for Toshiba America Business Solutions and Toshiba Global Commerce Solutions, technology leaders should become more visible to the business and lead by example to their teams. “I started attending the business meetings of all the other C-level executives on a monthly basis to make sure I’m getting the voice of the business,” he says. “Where are we heading? How are we making money? How can I help business leaders overcome their challenges and meet their objectives?”

In 2025, IT leaders should invest in AI, but also focus on the cases where they can demonstrate measurable value, and then improve on those cases incrementally. “Making huge investment decisions on AI very early can lead you into that trough of disillusionment, and that’ll be hard to pull out of,” warns Fernandes. “Make mid-scale investments and show ROI both in the short and long term, just like you would on any other project.”

What works for USF is to allow the business to decide which projects to implement, and make sure they share the risk. Fernandes says his team has made it a point to only invest where the business also invests to avoid a black hole of IT spending.

Prepare for general use of AI

Vendors are integrating AI into their most popular applications. Users who have depended on traditional packages for years need to be prepared for the change. This means not only learning about prompt engineering, but also remaining skeptical about some of the responses. After all, hallucinations won’t go away any time soon.

AI-empowered enterprise applications will change the way people work. According to Fernandes, IT leaders need to ensure both their staff and the business workforce are ready to do things differently to take advantage of the co-pilots. “This is going to be critical,” he says.

CIOs should also build platforms for custom tools that meet the specific needs not only of their industry and geography, but of their company — and even for specific divisions. AI models will be developed differently for different industries, and different data will be used to train for the healthcare industry than for logistics, for example. Each company has its own way of doing business and its own data sets. And within a company, marketing will use different data than customer service.

According to Guerrier, one of the most surprising things in 2024 was many IT leaders went ahead with AI without truly understanding their data topology — how data is taken in and formatted, and how it’s fed into an AI model. “It’s the same adage: garbage in, garbage out,” he says. “You can have the best AI tool, but if your data is ingested from a bad source, you’ll have bad outcomes from AI. IT leaders need to do a better job of managing their data in 2025.”

Fernandes says that IT leaders also need to secure data and IP, especially as agentic AI becomes more prevalent. “Agentic AI will have knowledge of the data in your data lake, which means your data governance, your loss prevention policies, and your cybersecurity processes have to be even stronger because you’re now going to expose data at a rate you can’t control,” he says.

Because most IT departments were caught off guard by gen AI, they didn’t have the skills they needed to run projects. To prepare for more challenges next year, IT leaders have three choices. They can build the skills in house, hire from outside, or develop strategic partners with trustworthy companies that have the skills.

Toshiba’s Roberge is creating an innovation and strategy department for the IT organizations he heads. “We’re going to identify and hire data engineers and data scientists from within and beyond our organization — and we’re going to get ahead,” he says.

Remember to keep the lights on

With all the commotion around gen AI, many IT leaders aren’t spending enough time on the other things they should be doing. “The emphasis on efficient operations needs to come back,” says Prasad. “At some point, it’ll come back to you if you lose sight of the fundamentals of your job.”

IT leaders should never forget that keeping the lights on (KTLO) remains a foundation of success, says Neal Sample, SVP and CIO of Walgreens Boots Alliance, Inc. While AI may be an exciting proposition, KTLO often offers a more certain payoff. Modernizing systems, consolidating platforms, and retiring obsolete solutions reduce complexity and create a more agile environment. “These steps not only lower costs and improve productivity, but also make IT more capable of supporting priorities like AI,” says Sample.

For example, AI thrives on clean, reliable data, making traditional IT practices like data governance and integration indispensable. Poor-quality data undermines even the best AI models, reinforcing the importance of foundational IT work. “Stable IT operations ensure that AI-embedded tools work as intended, minimizing disruptions and preserving trust,” says Sample. “The exciting potential of AI simply can’t be realized without the solid groundwork KTLO provides.”

According to Sample, IT leaders should also clean up technical debt to manage the complexity of modern IT environments. Outdated systems, overly customized applications, and fragmented architectures slow progress, increase risks, and make scaling innovations harder. Treating technical debt as a continuous investment keeps IT resilient and adaptable, ready to meet future challenges.

“While AI grabs the headlines, KTLO delivers steady, measurable results,” says Sample. “Investing in strong operational foundations and reducing technical debt creates the conditions for innovation to succeed. By focusing on these boring but essential priorities, CIOs can simplify their IT landscape and confidently drive progress in 2025 and beyond.”