From our work with enterprise clients across regulated and technology-intensive sectors, one thing is increasingly clear: generic GenAI experimentation is over. What matters now is specialisation, verifiability, and agent-based execution.
This article presents forward-looking predictions. Not every path is fully proven yet, but the signals from research publications, vendor roadmaps, and early enterprise implementations point in a consistent direction.
Key takeaways
- AI development is moving from general-purpose models towards vertical LLMs and specialised SLMs trained on domain-specific data.
- Verifiable reasoning frameworks, including reinforcement learning approaches tied to measurable outcomes - RLVR (Reinforcement Learning with Verifiable Rewards), are gaining importance in high-stakes environments.
- AI agents are evolving from assistants into embedded execution layers within enterprise workflows.
- Emerging standards such as Google’s Universal Commerce Protocol (UCP) signal the rise of agentic commerce.
- On-premise AI and infrastructure sovereignty are becoming strategically important for regulated industries.
- Competitive advantage increasingly depends on combining specialised AI models with organisational redesign and governance.
The end of the “one model fits all” era
In regulated industries such as finance, healthcare and science, relying solely on general AI models is becoming a strategic risk. The market is moving decisively towards:
- Vertical LLMs, trained on proprietary, domain-specific datasets
- Small Language Models (SLMs), optimised for narrow tasks and lower infrastructure costs
- Advanced reasoning frameworks such as RLVR (Reinforcement Learning with Verifiable Rewards)
The broader shift towards domain-specific AI systems is reflected in industry analyses such as the Stanford AI Index Report, which highlights rapid enterprise adoption and increasing focus on practical, domain-level impact rather than model size alone.
In healthcare and biology, the evolution of AI from pattern recognition to structured reasoning is visible in systems like DeepMind’s AlphaGenome, designed to improve understanding of genomic sequences and mutation effects.
Independent coverage in Nature further illustrates how such models may support research into rare diseases and biological mechanisms.
While it is too early to claim systemic clinical replacement, these developments demonstrate a clear trajectory: AI models are being engineered for domain reliability.
At the same time, SLMs allow organisations to extract smaller, industry-focused models that deliver high performance at a fraction of the infrastructure cost.
The conclusion is not that general models disappear. Rather, competitive differentiation increasingly comes from depth of domain integration, auditability, and alignment with regulatory constraints.
Get recommendations on how AI can be applied within your organisation.
Explore data-based opportunities to gain a competitive advantage.
From AI models to AI agents
Models are the brain, and in 2026, they have gained hands.
In 2026, AI systems are no longer confined to generating outputs, but they are increasingly embedded into operational layers across enterprise systems. They interact with APIs, orchestrate workflows, and trigger actions.
We can distinguish several emerging layers of agent maturity.
- Workflow agents – automating well-defined back-office processes.
- Orchestrated multi-agent systems – coordinating task-specific agents across complex value chains.
- Interface-controlling superagents – acting as unified entry points to multiple services and tools, significantly simplifying user experience while reducing licensing costs associated with fragmented software ecosystems.
- Physical-world agents – combining AI models with robotics platforms.
In robotics, Nvidia’s announcements around foundation models for generalist robotics illustrate how large-scale AI is increasingly integrated into physical systems.
AI systems embedded in robotics carry an additional implication: one of the key hypotheses in the development of Artificial General Intelligence (AGI) is the need to ground intelligence in real-world interaction. By enabling AI-powered robots to operate in physical environments, these systems can learn not only from abstract representations but also through direct engagement with reality.
These developments do not yet imply full autonomy across industries. They do however signal a structural shift: organisations are beginning to redesign processes around autonomous or semi-autonomous execution layers.
Industry discussions around AI agents and enterprise transformation are also reflected in analyses by major consultancies such as McKinsey and Gartner, which increasingly frame AI as an operating model transformation rather than a productivity add-on.
Agentic commerce and the end of the shopping basket
Google’s introduction of the Universal Commerce Protocol (UCP) signals a move towards standardised, machine-readable commerce interactions.
Additionally, industry coverage describes UCP as enabling AI agents to search, negotiate, and complete transactions on behalf of users.
If such standards mature and gain adoption, competition in e-commerce may gradually shift from interface design to technical accessibility for purchasing agents.
But this is still an evolving space. Regulatory and privacy concerns are already part of the public debate, as reflected in discussions around AI-driven checkout systems.
The long-term outcome is uncertain. However, the directional signal is clear: enterprises should prepare for machine-to-machine transaction environments where APIs, structured data and compliance design become strategic differentiators.
On-premise AI and infrastructure sovereignty
As geopolitical tensions and regulatory scrutiny intensify, infrastructure decisions are becoming strategic.
Local, on-premise AI deployments allow employees to manage files, knowledge bases and workflows without constant cloud dependency. The benefits are tangible:
- reduced latency in critical operations,
- greater control over intellectual property,
- compliance with strict confidentiality requirements.
For many regulated enterprises, local deployment is not a technical preference but a risk management decision.
The global AI landscape increasingly intertwines compute capacity, energy access and hardware sovereignty. Public discussions around large-scale AI infrastructure initiatives in the US and China highlight how compute ecosystems are becoming national strategic assets.
Geopatriation is not a transient trend, but a structural shift in how AI systems and IT infrastructure are designed. Gartner predicts that by 2030, more than 75% of enterprises in Europe and the Middle East will repatriate their virtual workloads into environments specifically designed to mitigate geopolitical risk, compared to less than 5% in 2025.
For enterprise leaders, vendor selection is therefore no longer only about model performance. It also involves long-term exposure to regulatory, trade and hardware dependencies.
Read more about AI on our blog:
The socio-economic impact: AI staffing and new roles
Another structural shift concerns workforce design.
Large enterprises are increasingly auditing processes to determine which functions can be automated, augmented, or fully “agentised”. Instead of simply reducing headcount, we observe the emergence of hybrid staffing models where autonomous systems operate under human supervision and governance.
According to LinkedIn trends, roles such as AI Consultant and AI Strategist are among the fastest growing. The key differentiator is no longer pure technical expertise, but the ability to combine domain knowledge with agent design and governance.
This transition is ongoing and uneven across industries. However, the direction is consistent: AI is moving from tool to organisational layer.
Strategic recommendations for 2026
Based on current signals and early enterprise implementations, several structural priorities emerge:
- Treat AI ecosystems as integrated operational layers, not isolated assistants.
- Prioritise stability and auditability in high-stakes processes.
- Invest in domain specialisation to create defensible differentiation.
- Conduct recurring process audits to identify agentisation potential.
- Define a clear infrastructure strategy, including on-premise and hybrid deployment options for strategic data.
Not all predictions outlined here will materialise at the same pace. Some may evolve differently due to regulation, market consolidation or technical bottlenecks. However, the strategic direction is increasingly visible: AI systems are becoming embedded, specialised and infrastructure-dependent.
What this means for business leaders
The coming phase of AI adoption is less about experimentation and more about architecture.
The organisations that succeed will not necessarily be those that experiment the most. They will be those that align specialised AI systems, agent-based execution and governance frameworks with clearly defined business outcomes.
AI can deliver unprecedented scale and speed. Competitive advantage, however, will continue to depend on strategic clarity, disciplined implementation, and organisational redesign.
Developing an AI platform that saves law firms up to 75% of document review time
