I’m trying to understand where AI is heading. In your experience, are industry-focused large models the future, or will flexible AI agents play the bigger role?
In my opinion, the real challenge is not choosing between AI agents or industry-specific large models, but designing the right architecture. Current LLMs can support both approaches, but without a deterministic and domain-driven structure, they remain unreliable. In my latest preprint, I propose a semantic and token-based architecture for vertical AGI in energy systems, where models are replaceable but the architecture ensures reliability, traceability and user control.
In my opinion, the future of digital transformation will be shaped by both AI agents and industry-specific large models, but in different ways. Industry-focused models bring depth, accuracy, and trust, making them vital in regulated fields like healthcare, finance, and law. AI agents, on the other hand, offer adaptability and automation, acting like digital employees that can plan, reason, and execute tasks across systems. The most promising direction is their fusion specialized models providing domain expertise while agents put that knowledge into action, enabling smarter, end-to-end automation. Together, they will drive the next major wave of transformation.
I believe both will have their roles in the transformation in different fields. While large models are very effective and efficient for commercial purposes, agent-based models are considered highly valuable for policy development and evaluation, providing insights to the public administration bodies
In my view, both industry-specific large models and flexible AI agents are critical, but they serve different purposes and will likely coexist in the digital transformation landscape. However, if I had to lean toward one as more promising for the future, I’d place my bet on AI agents—with an important caveat.
Industry-specific large models—think legal, healthcare, or manufacturing LLMs fine-tuned on domain data—are incredibly powerful for tasks like document summarization, compliance checks, or clinical note analysis. They bring deep contextual understanding and can dramatically improve efficiency within a vertical. But they’re still largely passive: you ask, they respond. They don’t act autonomously.
AI agents, on the other hand, represent the next evolutionary step. They don’t just answer questions—they reason, plan, act, and interact with systems on your behalf. Imagine an agent that doesn’t just tell you a supply chain is at risk but proactively reorders inventory, negotiates with vendors, and updates your ERP system—all while learning from feedback. That’s transformative.
That said, the most effective agents will almost certainly leverage industry-specific models under the hood. A medical AI agent, for example, will need a healthcare-tuned foundation to make safe, accurate decisions. So rather than an “either/or,” think of it as agents as the orchestrators, powered by specialized models as their domain experts.
In short: Industry models provide the depth; agents provide the agency. For true digital transformation—where systems don’t just inform but operate intelligently—agents are the more promising frontier. But they’ll thrive only when grounded in high-quality, domain-aware models.