In 2025, agentic artificial intelligence (AI) took over enterprise IT, and from late 2022 onwards, so will generative AI (GenAI). Behind both are years of traditional AI, based on machine learning, advanced statistical analysis, predictive modeling, and more.
What will 2026 hold? The most likely theory is that all forms of artificial intelligence (AI) will further combine and converge, and that companies will need to shape their governance around how that convergence will shape.
The question of whether the AI bubble in global stock markets will burst is an interesting question, but one that is less important to enterprise IT in user organizations. The dot-com boom and bust have come and gone, but the Internet has endured, changing every aspect of economic, social, and political life. So it's likely to happen with the rapid growth of AI.
One of the big questions that agent AI specifically posed in 2025 was around how organizations manage the identities of their agents. Managing the identity of human workers and machine workloads is one thing enough, but autonomous AI systems have proven to be particularly challenging.
An Okta study published in August 2025 shows the emergence and proliferation of new security issues related to the proliferation of AI agents and non-human identities. The focus on this topic, and the alignment of identity security and agent AI security, appears to be paying off by the end of the year, as seen in Q3 earnings.
However, many other suppliers are turning their attention to the impact of agent AI on cyber security, from both a defender and attacker perspective. ServiceNow's acquisition of AI-first identity management platform Veza and agreement to acquire Israeli AI security company Armis, announced in December 2025, illustrate this trend. And that's just one development.
As agentic AI becomes more secure, the technology could become widespread in 2026, delivering even more of its vaunted benefits of increased economic productivity and making employees' work more creative and enjoyable. The reason this is latent rather than definitive is that power within organizations is always contested and is not an abstract, purely technological issue.
The AI Hype Cycle as the Arrival of the Internet 2.0
However, resolving the cyber security issues that can plague enterprise applications is tactical in nature rather than strategic for businesses and other organizations.
The question of how we can productively compare and contrast the dramatically increased hype of AI with the emergence of generative AI in 2022, and its maturation into agent AI in 2025, with the rise of the Internet in the late 1990s is a question that requires more strategic consideration. This is also a plausible guide to what will happen next in 2026.
In a strong sense, we've been here before. The answer to this comparison question could suggest the direction of AI development in enterprises in 2026. Computer Weekly posed the question of comparing these two phenomena in the history of technology to a set of major IT industry executives during the 2025 Q4 conference season.
Steve Miranda, Oracle's executive vice president of Oracle Application Development, said ahead of Oracle AI World that he would not single out customer experience (CX) as the most promising area for AI development. “Back in 1998, our slogan at Oracle was 'The Internet Changes Everything.' If you remember back then, there was a huge focus on e-commerce.”
“Sure, e-commerce existed, but the Internet changed music and changed the way we listened to music. The Internet changed the way we watched. I'm not even talking about watching TV or anything. The Internet changed communication and meetings. It changed everything. The Internet changed everything.”
“Right now, we believe that AI is going to change everything. There are charismatic use cases right now, like customer service and CX. But when I say AI will change everything, I'm giving the example of invoice scanning and paying, and invoice scanning and paying. [a company’s] Ledger – I don’t think anyone else understands the magnitude of the overall impact this has. We are currently at the same stage as the Internet, but faster. ”
At SAP TechEd in Berlin, SAP Chief Technology Officer (CTO) Philipp Herzig said that how enterprise applications are developed will be more interesting than how the infrastructure or large-scale language models (LLMs) themselves are developed. He said the current rise of AI is moving much faster than the internet, adding, “But I still observe that it's comparable to the early days of the internet, when there was also a lot of focus on the infrastructure plans of companies that built switches, routers, etc.”
“The discussion then turned to ISPs. [internet service providers] – AOL and carriers [companies] It brought the internet into homes and built distribution to consumers. But the reality is that the value gained from cloud services is what we all consume today with streaming services and cloud software.
“For example, even though there's a lot of discussion going on around the hardware side and the power of data centers and GPUs, it's our belief that the end game is not going to be played there. The end game is going to be played with great user experiences and AI-native software as a service that people can just turn on. We're going to see a lot of new business models come out of that.”
It was in Berlin in early November. Later that month, Gerrit Kazmaier, Workday's president of products and technology, said this at Workday Rising EMEA in Barcelona. “The important thing to understand is that technology goes through cycles, and the first thing that always happens is infrastructure innovation. We saw this with Broadcom and other broadband internet. Cloud infrastructure built on the availability of bandwidth and the economics of making it possible to do business over the internet eventually gave rise to SaaS.” [software as a service].
“We're looking at the same cycle. Now we're in the second phase. We've created a very powerful computer infrastructure, and now we have a form factor for large-scale language models that can basically be leveraged to do another kind of computation, which you might call computational inference.
“This didn't exist before, and software vendors who fundamentally solve business problems can now leverage this to change the way business systems are built. Business processes are largely designed around the constraints of human reasoning…But with the advent of computational reasoning, we have an opportunity to completely rethink it.”
David Richardson, vice president of AgentCore AWS, told Computer Weekly at ReInvent in Las Vegas at the end of the year: “I have long-term optimism. This is important because there are going to be bubbles, ups and downs. The beauty of the Internet is that it allows for incredibly flexible communication between a wide variety of computing capabilities, and people have built things on top of it that no one could have predicted.”
“The first time I used the Internet was in 1987, seven years before the Web, but it was already fascinating. Another thing that happened with the Internet was that it created a massive ecosystem around building the Internet…[And now] There are so many different participants in the ecosystem who see opportunities to deliver value through innovation, and it feels like these models are years away from increasing reasoning and resiliency.
“We keep finding new ways to apply them to problems that we previously thought we couldn't deal with or couldn't deal with economically. So this is my long-standing outlook on what's going to happen. Twelve months from now, I have no clue and I'm not going to make any predictions about that.”
What does a basic comparison between the AI Hype Cycle and the early 20th century Internet's dawn tell us about the likely path for the development of AI-enabled business applications in 2026 and beyond? A new stage is poised to begin. Classical AI (or simply machine learning) will be joined by generative and agent AI, and enterprise applications will function more like video streaming services that algorithmically deliver content to consumers, not like the old broadcast TV that Gen X grew up with.
Claus Jepsen, CTO at Unit4, suggestively calls this “ambient ERP” and his predictions for AI in ERP in 2026 are interesting reading. Similarly, Stephen Webb, Capgemini's UK Chief Technology and Innovation Officer, makes a convincing argument in Computer Weekly's 2026 Forecast Opinion: “AI is no longer a supporting technology; it is rapidly becoming the operating structure of modern enterprises. We are seeing a decisive shift beyond single-tasking activities to autonomous, adaptive, self-optimizing systems leveraging multi-agent systems.”
Achieving that level of autonomy will require technology and governance processes that have been in development since late 2022, when generative AI first appeared, but will need to catch up. Michelle Eisenberg, General Counsel at Unit4, makes some interesting points about building an AI governance framework and balancing prudence and innovation beyond supplier origins.
But don't be fooled into thinking, “This is different. It's like we all woke up with 20 more IQ points,” as DXC Technology's Brad Novak, who has been through waves like “mobile, social, cloud, and big data,” told Computer Weekly at the 2025 Boomi World Tour in London.
