Agentic AI: "Human in the Loop" Is Too Small

A leader I talked to yesterday wasn’t sure what their work will look like a year from now. We were finishing an AI coaching sprint in which they shifted from chatting with AI and using Claude Code for very specific tasks, to unleashing agentic AI on several complex problems. Scaling their context. Integrating an AI second brain.

All great, but the reality of the AI doing more brings almost philosophical questions about “what’s remaining for me as a leader to do?”

“Human in the loop” is too small. The real shift is humans moving up a level. Agentic AI CAN unlock human agency.

We talked about elevating to more strategic loops and interactions. Shaping the work, the workflow, the agents, the guardrails, the next decision. Orchestrating and curating rules, values, skills across their team and across the company. Having the time and opportunity to ascend to a higher level of mastery of their field.

As I was reflecting on this conversation, Dan Pink’s autonomy / mastery / purpose model came to mind. But I’d probably swap autonomy for agency.

And mastery starts to look like this weird new craft: getting good at creating AI agents that make your old work disappear. Not so you become irrelevant. So you can take on bigger problems and opportunities.

I created an exercise way way way back that feels relevant again: the Autonomy Mastery Purpose retrospective. One interesting way you can leverage it as you’re thinking about AI adoption in your organization is to ask:

• How is what we’re doing with AI contributing to human agency? How are we in the way of it?
• What are we doing to create the conditions for people to pursue mastery? Where are we keeping people fixed to where they are, or even regressing in their skills and expertise?
• Are we creating a high-purpose environment? Or are we growing people like mushrooms (feeding them crap and keeping them in the dark)?

And for the fun of it, you can also run this exercise to explore what level of agency, pursuit of mastery, and connection to purpose you are giving your agents.

I often see “human in the loop” (HITL) being used as a “crowd control” device, promising employees that they will still be in the loop even when the company adopts agentic AI. Sometimes it’s a safety concern. The fear is that on its own, AI can make costly mistakes.

Most of the time HITL means human in the execution/workflow loop: reviewing, refining, or approving AI output. I’d argue you can’t 10x with this sort of HITL.

In order to 10x you need a different vision and focus for HITL. You need a human in the learning loop (aka double loop, or meta loop). You want human judgement, skill, and experience being used not just tactically but strategically. Observing the results of AI/human interaction in execution. Guiding the evolution of AI agents, agentic workflows, and even the lowly AI chatbots. So that over time there will be less and less need for humans in the execution loop.

This might take time. Humans might be needed in your execution loop for the foreseeable future. But the more effective AI is at helping them, the less capacity these humans need to spend on “running the business”, the more time they have to dedicate to “evolving/developing the business”.

Humans evolved to be tinkerers. To innovate. To try, to sense, to adapt. To continuously learn and improve.

That’s the loop we really belong in.