Models, Infrastructure, and Enterprise Readiness for Agentic AI - with Alex Tyrrell of Wolters Kluwer

· Source: The AI in Business Podcast · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Cloud Computing & IT Infrastructure, Cybersecurity & Data Privacy · Depth: Intermediate, long

Summary

Alex Tyrrell, SVP and CTO of Health at Wolters Kluwer, highlights that infrastructure readiness, not model performance, is the primary bottleneck for agentic AI adoption in healthcare. He distinguishes agentic AI's "do" capabilities—acting, planning, and executing tasks across complex workflows—from generative AI's "create" functions. Tyrrell emphasizes the need for domain-adapted reasoning, granular APIs, and enhanced observability as agents drive higher-volume system interactions. He advises preparing backend systems for increased load, adapting models to real-world workflows through techniques like fine-tuning and dynamic chain of thought, and avoiding monolithic architectures. The discussion also covers the evolving "build vs. buy" decision, favoring managed services for commodity tasks while advocating internal development for differentiated agentic workflows and specialized small language models (SLMs).

Key takeaway

For AI Architects and Directors of ML evaluating agentic AI for healthcare, prioritize backend infrastructure and modular system design over raw model performance. Your existing APIs and microservices must handle dramatically higher, agent-driven transaction volumes, requiring granular entitlements and robust observability. Avoid monolithic systems; instead, focus on domain-adapting foundational models and building specialized small language models for differentiated workflows to ensure secure, compliant, and scalable deployment within regulated environments.

Key insights

Agentic AI shifts focus from content generation to task execution, demanding robust infrastructure and domain adaptation.

Principles

Method

Domain-adapt foundational models using instruction fine-tuning, supervised fine-tuning, low-rank approximation, parameter-efficient fine-tuning, in-context learning, and dynamic chain of thought.

In practice

Topics

Best for: CTO, VP of Engineering/Data, MLOps Engineer, Director of AI/ML, AI Architect, Consultant

Related on AIssential

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by The AI in Business Podcast.