Governing at AI Speed

With a clear framework for calibrating guardrails to real-world stakes, the next question becomes: how fast can your organization actually update those guardrails as AI capabilities shift? The panel made clear that getting the risk levels right today means very little if your governance structure only revisits them on an annual cycle. This unit unpacks why traditional planning horizons are fundamentally mismatched with the pace of AI and what cross-functional, speed-matched governance actually looks like in practice.

The panel surfaced a striking mismatch: "If a higher education university is building their curriculum and deciding every five years or ten years, but the talent that students need is changing every three months, how does that align?" That example wasn't only about universities — it was a mirror for every organization still running governance on annual cycles. Taylor Stockton argued:

Stockton: "those processes, those decision-making approaches need to be fundamentally redefined such that they can take in what's happening and then act on it in a different way."

For HR leaders, this means advocating for quarterly or even continuous governance review cycles that match the cadence of AI capability shifts.

Building on this urgency, the panel also modeled who should be at the governance table. One organization described how the HR and IT departments co-sponsor AI governance through shared principles — anchored by a foundational commitment to "treat AI not as a tool but as an enterprise capability." That distinction matters. When AI is treated as a departmental side tool, governance fragments across silos. When it's elevated to an enterprise capability, every leader shares ownership of the guardrails, the escalation paths, and the speed at which they're updated. If you take one structural change away from this lesson, it's this: co-sponsorship between HR and IT isn't optional — it's the foundation that makes everything else work.

The discussion also addressed one of the most consequential governance questions emerging right now: who is accountable when AI agents start coordinating with each other? Jason Desentz argued you should treat agents like employees — assigning KPIs and monitoring performance just as you would a human. But Stockton pushed back, cautioning against how quickly that framing can escalate:

Stockton: "that gets you down a rabbit hole to say, well, you can just have an entire org chart of agents here and they all coordinate together."

The real risk, Stockton argued, is losing sight of who actually owns the decisions:

Stockton: "not having humans in the right place, such that you have that accountability to say what human is ultimately responsible for each of these decisions."

Particularly with agent-to-agent interactions already happening, the panel emphasized that the temptation to scale autonomous agents must be met with clear human ownership defined before deployment — not discovered after something goes wrong. The discipline here is simple but hard: for every agent you deploy, name the human who owns it.