I haven’t sailed in over a month, so I’m running out of sailing analogies—making this an AI-only post without any marine analogies.
Google’s General Counsel recently said, “I view it [AI] as a complement, not a replacement.”
I agree with that sentiment, but the advent of autonomous agents is making it harder for humans to exercise judgment. Almost nobody argues that humans should disappear from consequential enterprise decisions. We hear a lot about human-in-the-loop, human oversight, responsible AI, copilots, and agent guardrails.
But most of these concepts focus on controlling AI. The problem I think is critical is how to make human judgment part of AI reasoning.
Here’s an example enterprise task:
“Cut the wholesalers below the 12% gross-margin floor.”
The business expert responds:
“Don’t cut six of them. They’re in hurricane-affected territories. The disruption is temporary.”
That’s not an approval or a guardrail. It’s human judgment changing the AI reasoning.
Google’s General Counsel is right. Human judgment isn’t going away. But keeping humans around isn’t enough. Human judgment has to become part of the organization’s intelligence—not disappear when the AI conversation ends.
My thesis is that humans should not just be in the loop—a very passive role—but should share responsibility for reasoning with AI agents. Further, both human and AI reasoning should be captured in systems as Decision Intelligence.
Supplying context to AI models is necessary but not sufficient. The interaction between AI models and humans, along with existing enterprise assets, must be two-way. That implies interoperability between AI agents and existing enterprise systems, data, and human expertise.
That’s an interoperability problem I think Enterprise AI needs to solve.