Sotiris Leventis is an Enterprise Architect and AI specialist with over 25 years of experience designing intelligent systems and enterprise software solutions. He is the Founder and CEO of Qognity, where he leads the development of multi-agent systems.
His AI journey began with academic research in neural networks and agent-based systems, later evolving into hands-on development of platforms for infrastructure monitoring, legal tech, and enterprise automation. He has also contributed to publications on AI applications in governance and law. Previously, he founded Hypernetica, delivering solutions to EU institutions through strategic partnerships, and held leadership roles at Microsoft and other technology companies.
He is a recipient of Microsoft’s Great People Great Performance Award, a MENSA awardee, and has mentored students and startups through Google Summer of Code and Microsoft Startup Weekend. Sotiris also serves as Head of Technology for the Hellenic OCR Team, supporting civic innovation through AI.
Before “AI agents” were the trend everyone claimed to have invented last Tuesday, the 1990s delivered the first serious wave of multi-agent systems.
These “Millennial Agents” grew up in a world explicit protocols, & architectures so well defined you could almost hear them ask for a design review meeting. They brought structure & discipline, but required rigorous work to make even modest autonomy work.
Fast-forward to the 2020s, where “Gen Z Agents” burst onto the scene with the confidence of knowing every answer, even the ones they occasionally make up. Powered by LLMs, they can reason broadly, talk fluently & plug into tools in ways the 90s only dreamed of. But many deployments still rely on a single, all-purpose agent. The demos are dazzling, the production systems…sometimes less so.
This talk contrasts the thoughtful, architecture-loving Millennial agents with the fast, flexible, sometimes overly optimistic Gen Z LLM agents. Drawing on Qognity’s real-world enterprise deployments, we explore how to blend the best traits of both generations: the rigor, specialization & coordination the 90s showcased, combined with the adaptive reasoning, retrieval & tool fluency LLM agents excel at.
Attendees will learn what it takes to move beyond hype-driven agents & instead build robust, well-governed multi-agent systems, that work not because they are trendy, but because they combine the strengths of two very different & surprisingly complementary, generations of agent thinking.
Searching for speaker images...