- The 70+ best Black Friday TV deals 2024: Save up to $2,000
- This AI image generator that went viral for its realistic images gets a major upgrade
- One of the best cheap Android phones I've tested is not a Motorola or Samsung
- The best VPN services for iPhone: Expert tested and reviewed
- Docker Desktop 4.36 | Docker
CIOs to spend ambitiously on AI in 2025 — and beyond
While the ROI of any given AI project remains uncertain, one thing is becoming clear: CIOs will be spending a whole lot more on the technology in the years ahead.
Research firm IDC projects worldwide spending on technology to support AI strategies will reach $337 billion in 2025 — and more than double to $749 billion by 2028.
Those bullish numbers don’t surprise many CIOs, as IT leaders from nearly every vertical are rolling out generative AI proofs of concept, with some already in production.
Nate Melby, CIO of Dairyland Power Cooperative, says the Midwestern utility has been churning out large language models (LLMs) that not only automate document summarization but also help manage power grids during storms, for example.
Global professional services firm Marsh McLennan has roughly 40 gen AI applications in production, and CIO Paul Beswick expects the number to soar as demonstrated efficiencies and profit-making innovations sell the C-suite.
JP Morgan Chase has also made myriad generative AI investments in its investments businesses as well as its Chase Travel, contact center, operations center, and credit card services bureau.
“Generative AI is a transformative technology and the focus that we have today will be doing use case-based approaches,” Chase CIO Gill Haus says, noting that he is not worried that the ROI will disappoint.
AI spending on the rise
Two-thirds (67%) of projected AI spending in 2025 will come from enterprises embedding AI capabilities into core business operations, IDC claims.
Some will grab the low-hanging fruit offered by SaaS vendors such as Salesforce and ServiceNow, while others will go deep into laying the enterprise infrastructure for a major corporate pivot to AI.
The big investments in generative AI may eventually rival traditional cloud investments but that does not mean top cloud providers — all of whom are top AI platforms providers — will suffer. Amazon Web Services, Microsoft Azure, and Google Cloud Platform are enabling the massive amount of gen AI experimentation and planned deployment of AI next year, IDC points out.
Cloud providers offer most organizations the least risky way to get started with AI, as they do not require upfront investments or long-term commitments. Also, cloud providers offer the latest AI advances, whether that’s the underlying GPU infrastructure or the developer platforms needed to build AI applications, says Dave McCarthy,
vice president of cloud and edge infrastructure services at IDC and one of several analysts who conducted IDC’s research.
“Enterprises are also choosing cloud for AI to leverage the ecosystem of partnerships,” McCarthy notes. “Cloud providers have become the one-stop shop for everything an enterprise needs to get started with AI and scale as demand increases.”
In fact, the two technological advancements are fully symbiotic, McCarthy points out.
“The emergence of AI, especially generative AI, further amplifies the potential of cloud computing, enabling organizations to boost productivity and explore innovative business models,” he says.
IDC also surveyed IT leaders on their build vs. buy equations for AI. A third (34%) plan to use AI capabilities built into existing enterprise applications, such as Microsoft Copilot in its Office suite and Google Gemini in its Workspace, McCarthy says.
“This also extends to SaaS providers like SAP and Salesforce that are adding AI features to their products,” he says. “This is the easiest way to start benefiting from AI without needed the skills to develop your own models and applications.”
According to IDC, 53% of enterprises plan to start with a pretrained model and augment it with enterprise data. Only 13% plan to build a model from scratch.
No matter what route CIOs take, it’s clear the genie is never going back in the bottle.
“In the near term, most enterprises are focusing on automation and productivity use cases that can be implemented without fundamentally changing business processes,” McCarthy adds. “However, the higher value use cases involve new business models, which require widespread organizational change.”
Stephen Crowley, senior advisor for S&L Ventures and former CIO of global technology solutions at Covetrus, still sees that future as a little way off.
“Building the foundation is different from moving to production with AI apps. I think that will take longer,” he says. “Even so, there will be tremendous spend on foundational capabilities and those who provide those capabilities such as cloud providers will do very, very well.”
Sound foundations, good governance
Marsh McLennan’s Beswick says the firm will continue its aggressive embrace of gen AI to move beyond basic applications and automate internal business processes.
“It’s shifting the automation frontier and many of those opportunities are worth maybe thousands and sometimes millions of dollars,” he says, adding that building a platform first and using models based on OpenAI has allowed Marsh McLennan to put gen AI to work far less expensively than many think.
“The thing about the AI stuff is it’s really cheap, if you do it right,” Beswick says. “It’s really just keeping up with the rate the technology is moving, constantly challenging the assumptions of what we’ve built so far.”
For the global risk advisor and insurance broker that includes use cases for drafting emails and documents, coding, translation, and client research. Beswick estimates as many as a million hours have been saved by these and other homegrown gen AI innovations, including data schema extractors, RFP first-draft generators, and models that analyze hundreds of thousands of data points using natural language at scale.
“We built this whole thing so that the models are swappable, and we’ll constantly be evaluating which models we use based on mostly price performance, but also just what the risk profile looks like to us,” says Beswick, whose team also built a security and governance platform to serve as the foundation for the firm’s AI development. The firm has also established an AI academy to train all its employees.
“We anticipate using models from Amazon and Google and possibly others as part of the mix as we move into next year. We think there’s a lot of value in small language models that are fine-tuned for specific use cases” as well, he says.
CIOs, like Beswick, are also forming in-house AI committees and establishing AI governance rules designed to prevent damage to their enterprise and to ensure unchecked “shadow AI” is kept at a minimum.
Chase’s Haus, for example, firmly believes that the intense amount of work currently invested in fixing gen AI hallucinations and setting up guardrails and governance will pay off substantially over time.
The risks of accidentally exposing sensitive corporate data or designing gen AI models that fly afield of their intended missions are also top of mind for Dairyland Power’s Melby, who is working with a Microsoft partner to deploy Copilot and Azure OpenAI capabilities to employees in a secure manner.
“CIOs need to be aware of the changing tides in this space and meter investment to align with the risk tolerance of their company,” Melby says. “It becomes a cost vs. benefit conversation, where there can be huge advantages if the company is willing to adjust as more governance becomes established.”