A seminar by Marco van Hurne

*AI Factory leader at Eigenvector Research*

LLMs are great at language. Enterprises, inconveniently, run on processes, decisions, controls, systems, and coordination.

In this seminar, I’ll explain **why LLM-first architectures** struggle with complex workflows, why reliability falls apart across many steps, and why adding more agents, guardrails, and validators often creates a larger pile of technology rather than a better solution.

We’ll also explore **what comes next**: specialized models, deterministic execution, enterprise memory, ontologies, and coordination models.

*Not an anti-LLM rant.*

More of an architectural intervention before autocomplete gets promoted to CEO.

We’ll break down:

⚙️ Why multi-step LLM workflows break

🧠 Specialized models vs. generalist LLMs

🏗️ Deterministic execution, ontologies, and enterprise memory

📍 **Live Webinar 21 August 2026 at 15:00 CEST**

🔗 **Register here:** https://attendee.gotowebinar.com/register/8131101150082168416?source=MU

Description supplied by the event organiser.