It took the telephone 50 years to reach 50 million people, but according to Deloitte’s Tech Trends 2026 report, generative AI gained its first 100 million users in just two months.
However, according to the report, much of the current AI activity within enterprises and organizations remains stuck in the pilot phase: 38% of organizations are piloting agentic AI, but only 11% have transitioned those pilots into production.
As organizations push their AI efforts forward in 2026, they must focus first on the business case, stresses Kelly Raskovich, executive editor of Tech Trends.
“You need to make sure that you lead with need, and not the technology,” she said.
“AI will continue to be a big value driver, but people are on a journey with it. They need to make sure that they’re implementing the right use cases, and that they’re solving the right problems.”
Deloitte’s Tech Trends 2026 report focused on five aspects of AI to watch in 2026, including how AI will interact more broadly with the physical world, how agentic AI will drive IT hardware and software modernization and how AI will lead to more advanced data security measures.
AI Goes Physical
Deloitte defines physical AI as systems that enable machines to autonomously perceive, understand, reason about, and interact with the physical world in real time. These include not only robots, but also sensors, simulation systems, and autonomous vehicles. Unlike traditional robots, physical AI systems perceive their environment and adapt their behavior based on data, rather than merely following preprogrammed instructions.
Advances in computer vision, battery life, spatial computing, and sensors have all made physical AI more accessible and capable, the report notes.
“Amazon just deployed their millionth robot,” Raskovich said.
“Cincinnati uses drones to inspect bridges. They’re doing in minutes what previously took them months. The costs are decreasing, and physical AI is becoming easier to scale.”
Agentic AI Gets a Reality Check
Deloitte’s Tech Trends 2026 report notes that AI agents have captured the attention of enterprises, with the promise of automating existing processes and even serving as a “silicon-based workforce” that can power productivity alongside human employees.
However, the report stresses, success with AI agents will require leaders to rethink both their infrastructure and operations, with legacy systems, data architecture constraints, and governance and control frameworks all potentially acting as bottlenecks. The report frames this as an agentic AI “reality check,” citing a Gartner prediction that more than 40% of agentic AI projects will fail by 2027 because legacy systems cannot support the modern demands of AI execution.
Notably, 42% of organizations are still attempting to develop their agentic AI strategy, and 35% report having no agentic AI strategy in place.
Additionally, Raskovich noted, leaders will need to consider how existing processes can be redesigned to best leverage the capabilities of AI agents, rather than simply dropping the agents into existing workflows and expecting them to perform at a high level.
“I think people who are putting agents into existing processes might have a harder time than those that take a step back and say, ‘Let’s redesign this with agents in mind,” Raskovich said.
AI Infrastructure Goes Hybrid
Deloitte noted that when a significant portion of an organization’s IT infrastructure is hosted in the public cloud, there are frequent API calls and other factors that contribute to escalating costs, in addition to raising concerns about data sovereignty, latency, intellectual property protection, and resilience. Even though the industry has seen token costs drop dramatically in recent years, Deloitte notes that usage is rising faster than costs are falling.
“The solution isn’t simply moving workloads from cloud to on-premises or vice versa,” the report states. “Instead, it’s building infrastructure that leverages the right compute platform for each workload.”
In practice, this means using the cloud for variable workloads and bursting capacity, on-premises resources for predictable and continuous workloads, and the edge for uses where even small differences in latency can determine success or failure.
“It’s not one-size-fits-all anymore,” Raskovich said.
AI Reshapes IT Orgs
It’s not just infrastructure that needs to be rearchitected. Organizations are also making significant changes to their IT organizations to best utilize AI.
According to Deloitte, this change is likely to look far different from what many have envisioned. While the rise of AI has led to widespread fears about worker displacement, nearly 70% of tech leaders actually plan to grow their teams in response to AI.
“We’re seeing augmentation happen, and there are also new roles coming to light related to things like agentic AI and even physical AI,” said Raskovich.
“It’s about taking a step back and making sure that we’re evolving the right people, processes, tools, and technology, and that we can scale accordingly. I see it as an opportunity to redefine what success looks like.”
Security Takes Center Stage
Increasingly, organizations are taking a harder look at what new risks AI tools introduce and how to mitigate them.
Broadly, Deloitte notes, these risks can be categorized into data security risks (such as the danger that AI tools will inadvertently expose sensitive data) and AI model security risks (for instance, models gradually degrading over time or even gaining and using excessive authority to perform unintended actions). One particularly dangerous area is shadow IT, the unsanctioned deployment of AI tools by employees, which can create blind spots and expose sensitive data to training models.
Raskovich said organizations must proactively address these risks, not only to keep their data and systems safe, but to ensure that leaders and workers feel confident using AI tools.
“At the end of the day, trust trumps tech,” she said.
Calvin Hennick is a contributing writer. His work appears in BizTech, Engineering Inc., The Boston Globe Magazine and elsewhere. He is also the author of Once More to the Rodeo: A Memoir. Find him on LinkedIn.
© 2026 Nutanix, Inc. All rights reserved. For additional information and important legal disclaimers, please go here.
Related Articles
