Generative AI’s trial period is coming to an end. By 2026, the spotlight is shifting from models that summarise information to autonomous systems that act on it. The next year is expected to mark a clear break from today’s chatbot-driven landscape as organisations adopt AI agents capable of planning, reasoning, and completing complex tasks with little supervision.
Experts say this transition will push companies to rethink everything from infrastructure to energy use to how they measure productivity.
A move toward fully autonomous systems
Hanen Garcia, Chief Architect for Telecommunications at Red Hat, says 2025 was the year companies experimented with generative models. In 2026, he expects a “decisive pivot” toward agentic AI: software entities that can make decisions, carry out multi-step workflows, and respond to changing conditions on their own.
Telecommunications and heavy industry are becoming early proving grounds. Garcia points to progress toward autonomous network operations, where systems can configure and repair themselves. The goal is to reduce operating costs and shift value away from basic infrastructure and toward intelligence embedded within it.
Ryan Daws
This shift is also driving adoption of multiagent systems. Instead of relying on a single model, companies are deploying collections of specialised AI agents that can collaborate on industrial tasks. While that boosts capability, it also creates new security risks.
Security and energy become defining constraints
Emmet King, Founding Partner at J12 Ventures, warns that more autonomous systems mean more opportunities for hidden or malicious instructions to slip into workflows. He expects security efforts to move from traditional endpoint protection to monitoring and auditing AI behaviour itself.
Energy is another constraint that’s rising fast. As AI workloads scale, King argues the real bottleneck isn’t access to models but access to power. He believes grid capacity will shape which startups can scale in Europe, making energy policy a key part of AI strategy.
Sergio Gago, CTO at Cloudera, expects companies to adopt new performance metrics that prioritise efficiency over size. He predicts the competitive advantage will favour systems that use resources intelligently, not simply those trained on the largest models.
High-value workflows take the lead
Broad, general-purpose copilots are likely to struggle unless they deliver clear returns. Analysts expect the strongest results in manufacturing, logistics, and advanced engineering, where AI is woven into technical workflows rather than consumer interfaces. Buyers are becoming more demanding, and tools that can’t prove real productivity gains will lose ground.
Software becomes disposable
The way organisations build and use software is also set to change. Chris Royles, Field CTO for EMEA at Cloudera, says 2026 will usher in a more fluid model where users request temporary modules generated on the fly. These short-lived “disposable apps” can spin up and shut down in seconds, replacing traditional packaged software.
To support this shift, companies will need better visibility into how these modules are created and how they make decisions, ensuring errors are caught early.
Rethinking data storage and governance
With AI systems generating more data than ever, storage is reaching its limits. Wim Stoop, Director of Product Marketing at Cloudera, believes organisations will start discarding AI-generated content once it’s no longer needed. Verified, human-generated data will gain value, while synthetic data becomes temporary.
Stoop expects specialised governance agents to monitor data environments continuously. These automated “digital colleagues” could adjust permissions, secure sensitive information, and manage policy compliance without waiting for humans to intervene.
Sovereignty and the human element
Data sovereignty remains a priority for European companies. A Red Hat survey found that 92 percent of IT and AI leaders in EMEA consider open-source software key to maintaining control over infrastructure and data. Providers are likely to use existing data centres to deliver sovereign AI options that meet local regulatory requirements.
King believes control over training pipelines and energy supply will matter more than owning models themselves. Open-source progress is making it possible for more organisations to run advanced systems without relying on a few major providers.
As AI enters workplaces more deeply, human factors will matter too. Nick Blasi, Co-Founder of Personos, says tools that ignore personality, tone, and temperament will quickly feel outdated. By 2026, he predicts AI will flag a significant share of workplace conflicts before managers notice them. These systems will focus on communication, trust, motivation, and conflict resolution, bringing personality science into the core of enterprise AI.
PersonosA more grounded era ahead
The days of simple wrappers around existing models are fading. Organisations are starting to judge AI by measurable results, not branding or scale. As autonomous systems take hold, companies that control their data, energy, and workflows will have the strongest advantage.
