As generative AI moves out of its infant stage this year, it will begin to move through the pain of the early stages of a new technology and ultimately define it for the years to come, when agentic AI will dominate the industry and enterprises will focus on orchestrating agents.
In the last three years, Generative AI A technology that has evolved from multiple vendors and enterprises to software platforms with most technology vendors now providing an agentic capability or co-pilot.
The early years of the technology saw a flood of large language models, and it began to seem that the larger the parameter and context window sizes, the better. Technology shifted from larger models to smaller models with little emphasis on open source, including models from Chinese vendors DeepSeek and Alibaba.
Over the past year, the emergence of logic and thinking models has highlighted that markets have shifted from a focus on models to a focus on agents. Additionally, there is a significant new emphasis on the importance of data centers.
Continuous Acceleration and Experimentation
The rapid advancements in generative and agentic AI seen every year since OpenAI released ChatGPT in November 2022 are expected to continue into 2026. Many in the AI industry and outside observers are paying close attention to the state of the AI market, whether it is in a financial bubble and whether that bubble will burst anytime soon. This year, enterprises will look to make greater use of AI agents with clear purposes within their organizations and geopolitical locations and with clear uses in verticals such as media and retail.
“We’re going to look at 2026 as a year of acceleration,” said Futurum Group analyst Bradley Shimin. “Not in terms of investment because that has already happened, but in terms of optimizing the spend for data in the service of AI.”
He said decision makers will have to decide how to save money and what to cut.
Shimin said these options fall among several possible outcomes related to the AI boom-bubble question.
“You have this race that there will either be a pop of the bubble where everyone is disappointed, and ‘Oh my God, why did we invest in this?’ He said. “If that’s not reached earlier, we could see some of the bubble shrink back to normal… or you could see a sustained increase that is based on actual capacity.”
Shimin said the different options stem from the fact that generative AI technology is still a young technology.
“We’re still learning, we’re still understanding what this spacecraft is,” he said. “We find ourselves operating without really understanding what the thing does and how it actually works.”
cost counting
With most enterprises still in the experimentation phase and still conducting AI projects, 2026 could be the year where the push is to “get it right,” said Mark Becque, an analyst at Omdia, a division of Informa TechTarget.
“This is the year that AI continues its systematic journey toward practicality,” Becque said.
like enterprises work together Generative and agentic AIOne challenge, Becque said, is understanding the risks and making sure organizations can use AI tools to address a particular problem.
One way to reduce the risks of experimentation is to reduce the high costs associated with AI technology.
“There is pressure throughout the industry to reduce costs,” Becque said. “A big concern is that costs have to come down to make these things work.”
One expensive area is the data center.
“We will have a bubble, and things will go down if the cost per use or the cost of doing something doesn’t go down,” Becque said, referring to the idea that enterprises may use AI technology less if the cost of AI products and services remains high.
Better agentic AI technology, data and shadow agents
This will also be the year when model makers release more capable multimodal AI models that better support AI agents and enable greater orchestration of AI agents.
“Setting up and running a multi-vendor agent ecosystem that truly collaborates predictably and productively has been difficult,” said William McCann-White, an analyst at Forrester Research. He said that as interoperability advances, agent orchestration will become more successful.
“It’s still going to be a process of growing and learning,” McCown-White said. “Not everything is being created equal; some platforms will simply be unusable.”
McKeon-White said organizations will also need to learn how to build data environments that are more secure, with better permissions systems and optimized for their intended use so that AI agents or AI tools work as intended.
“A lot of investment organizations are engaged in improving their data environments, making them more secure,” he said. Without a healthy data environment, the improvements that AI model makers can make may be limited. Governance also needs to be improved to act as checks and balances on the models and ensure that they work correctly.
increase in shadow agents
Additionally, enterprises need to control the use of AI agents in their organizations and be aware of shadow agents, according to Suja Visvessan, IBM’s vice president of products.
shadow agent They are agents used by employees that the organization does not approve of.
“Enterprises need to keep an inventory of what’s happening,” Visvesvan said. He said shadow agents could also include organizations that know when an agent’s lifecycle has ended, with metrics and proper governance so they can stop it when an unapproved agent tries to access something it shouldn’t.
He added, “It’s on the enterprise to make sure they’re doing what they’re expected to do so they can have visibility into it.” “For all these agents and applications that you’re running, we need to know the lifecycle.”
Sovereign AI
Beyond the innovation emerging in agentic AI and new models, another outlook for 2026 is continued growth sovereign ai.
Sovereign AI is the idea of a nation controlling its own AI technology, such as infrastructure and software, to serve its interests.
According to Omdia’s Beque, the popularity and acceptance of sovereign AI will increase in the UK, EU countries and India in 2026.
“This will displace some sellers based in the US and China,” he said. “It won’t displace GPUs and CPUs because they can’t make them quickly enough.”
He said the sovereign movement will primarily impact multimodal AI models that are deeply rooted in local languages. For example, having AI models in a language other than English gives regional sellers a better chance to succeed. This allows enterprises in those regions to access better models tailored to their language. This will lead to more AI initiatives in production.
In India’s case, the market is big enough to compete with China and the US, Becque said.
“If the sovereign AI regulations there can boost the Indian ecosystem… then it gives them an opportunity to innovate, understand things, thrive there and then maybe expand outside India to challenge AI businesses,” he said.
More industries affected by AI
In addition to sovereign AI, more industries and verticals will continue to be impacted by AI technology in 2026, especially retail.
During the last holiday season, the number of shoppers visiting websites using AI chatbots before making a purchase increased.
That trend will continue this year, said Greg Zakowicz, e-commerce and retail consultant for marketing automation platform Omnisend.
“It’s going to be an accelerated evolution of buyers turning to AI as part of their research and discovery process,” Zackowicz said.
He said that in addition to AI platforms like ChatGPT that drive traffic to websites, the next opportunity for brands big and small are on-site powered chatbots like Walmart’s Sparky and Amazon’s Rufus.
“These chatbots provide a natural handoff from a larger AI platform to the on-site experience and can help buyers refine their search,” he said.
Additionally, it’s likely that by the end of the year, buyers will have full autonomy to go to another AI chatbot website like ChatGPT or Perplexity and buy whatever they want from there, Zackowicz said.
Another industry that will continue to see change is media. Some fear that AI technology will eliminate many media jobs. However, AI technology can also continue to help in the specific field of entertainment and media.
“AI will not replace creativity but will solve the metadata mess,” said Paul Pastor, founder and chief business officer of software media company QuickPlay. Over-the-top video distribution. As a cloud- and AI-agnostic company, Pastor said QuickPlay integrates technology from Gemini, TwelveLabs and AWS into its systems.
While the media and entertainment industries are using generative AI to dub long-form content and create clips, the next step is how to use these tools together.
“That’s what’s missing,” the priest said. “How they all work together to achieve efficiencies… (and) how to improve the entire ecosystem.”
According to Pastor, the need to tie together AI tools used for media and entertainment will lead to an increase in connectivity points between recommendation engines, ad-tech, CMS, and analytics.
“(AI) will become the connective tissue that connects disparate systems, extracting more value from current assets and providing more value to consumers by accurately predicting what the user wants and needs to look forward to, preventing them from wandering off to another platform or, worse, hitting the cancel button,” he said.
With all of these innovations expected in 2026, enterprises may feel overwhelmed with technology and how to use it most effectively. However, there is one thing enterprises can hold on to, said Futurum Group’s Shimin.
“The positive side of the use of AI has been shown and it is absolutely solid,” he said. “It’s just a matter of time, investment and effort, not only to maintain that value, but to grow it in directions that we don’t even know we can do, that we haven’t even thought about yet.”
