Guillaume de Saint-Marc, VP of Engineering at Outshift by Cisco, argues that multi-agent swarms represent a structural shift beyond single-agent AI. The core claim: systems where agents self-organize, delegate, and collaborate can tackle enterprise problems that no single model can solve alone. The discussion covers specific frameworks including Moltbook and OpenClaw, and introduces two concepts worth understanding before the industry catches up: the Internet of Agents and the Internet of Cognition.
The governance problem gets real attention here. Saint-Marc does not treat guardrails as an afterthought. At 12:02, the conversation turns to controlling agent behavior at scale, which is where most multi-agent deployments currently break down. IT operations is named as the first serious enterprise use case, a grounded claim backed by Cisco's infrastructure position rather than speculation.
The open source strategy segment at 20:49 is the sleeper section. How Cisco chooses to release agent tooling will shape which standards win. If you work in enterprise AI, agent infrastructure, or platform strategy, the full interview is worth your time. The self-forming systems discussion alone, starting at 08:58, reframes what agent orchestration could look like in 18 months.
[WATCH ON YOUTUBE →]