Every year, the tech industry races toward the next big thing. In 2025, that was unmistakably agentic AI, AI systems designed to act, decide, and automate workflows across the enterprise. As Beth Pariseau and I discussed on the Informa TechTarget – IT Ops Query podcast, 2026 won’t be defined by hype. It will be defined by accountability, and especially by cost and return on investment. We recorded this just before the end of the year, pre-CES, and more New Year’s news, so I will see how I do on my predictions.
My Predictions
• AI cost optimization becomes the #1 priority — ROI, not hype, will determine which AI projects survive
• Agentic “AI buddies” move into production — quietly modernizing finance, ops, and customer support
• Data silos persist, forcing new interoperability layers instead of one giant AI lake
• Data governance, security, and quality become mandatory as regulations and risk increase
• Boards slow hiring and push AI to expand workforce capacity rather than replace it
• AI resilience and cyber recovery become mission-critical as AI becomes embedded in core business systems
Let’s explore how AI cost optimization becomes the #1 priority. After a year of massive spending on GPUs, data platforms, and AI experimentation, enterprises are entering what I call the “optimization phase.” The question boards and CFOs are asking is no longer “Can we do AI?” but rather “What are we getting for what we’ve spent?”
From AI Gold Rush to ROI Reckoning
In 2025, companies poured millions, sometimes tens of millions, into AI infrastructure. We discussed examples such as buying 1,024 GPUs for a single AI “superpod,” often costing over $17 million. But the GPUs were just the beginning. You also need massive storage, networking, data pipelines, and often multiple copies of the same data just to keep the models fed.
What got overlooked? Cost optimization.
We’ve seen this movie before in IT. Data gets duplicated, siloed, copied into multiple clouds, and suddenly organizations are paying to store, govern, and secure the same data three or four times over. AI has poured gasoline on that fire. And in 2026, that bill comes due.
That’s why I believe cost optimization and ROI will be the single most important enterprise tech story of 2026. Not security. Not even new models. But whether organizations can show that AI is delivering measurable business value relative to what they’ve spent.
Why “Bigger AI” Will Fail
One of the biggest risks we discussed is that many companies are trying to do AI too big, too fast, attempting to automate entire supply chains or massive business processes in one leap. In my view, those projects are far more likely to stall or fail before reaching production.
AI is powerful, but it is not infallible. It struggles with context, depends heavily on the quality and availability of data, and still needs human oversight. When enterprises bet everything on one enormous AI transformation, they risk blowing their budgets without ever seeing meaningful returns.
In contrast, organizations that focus on targeted, high-ROI use cases, such as finance, customer service, and operations, will be the ones that succeed. AI “buddies” that help accountants close books faster or support teams resolve tickets more efficiently will quietly generate far more value than flashy moonshot projects.
Data Silos Are Now a Cost Problem
Another theme that ties directly to cost is data fragmentation. Most enterprises are not moving all their data into one magical AI-ready lake. Half of large organizations are still running core workloads on-premises, and that isn’t changing anytime soon. And this is the least of most organizations’ worries when it comes to their data platform strategy. (See below from Future of Data Platforms Summit Study, November 2025.)
That means AI has to operate across silos. But every extra copy of data, every new platform, and every new governance framework adds cost, risk, and complexity. In 2026, enterprises will push back against vendors that try to force everything into their proprietary stacks, because data gravity and regulatory realities make that economically unsustainable.
This is where open standards, metadata interchange, and better governance become not just technical necessities, but financial ones.
The Real Promise of AI in 2026
Despite all this, I’m not pessimistic. In fact, I’m more optimistic about AI than I’ve been in years.
2026 is when we’ll see whether agentic AI truly becomes an enabler rather than a science project. Boards are already signaling that hiring will slow, and organizations will be expected to “do more with what they have.” AI won’t replace people, but it will allow fewer people to handle larger scopes of work.
That is where the ROI lives—and where the cost savings emerge.
When AI helps a finance team close faster without adding headcount, or enables a support organization to handle more volume without burning out staff, that’s real economic value. That’s what will separate winners from companies that simply burned money on infrastructure.
