Skip to content

The Architecture of Commitment Promises, requests, and the network of conversations, Flores via Winograd/Flores

Draw your organization chart.

Boxes. Lines. Hierarchy. Reporting relationships. Departments. The visible structure of who is responsible for what. This is how we’re taught to see organizations as structures that can be mapped, optimized, reorganized when they fail.

Now throw it away.

Not because it’s wrong. Because it’s describing the skeleton while the living body is somewhere else entirely.

The living body of any organization, any team, any family, any collaboration of two people trying to accomplish something together, is a network of commitments. Open requests. Pending promises. Declarations that created new realities. Assessments that are shaping decisions without being spoken aloud. Conversations that completed cleanly and conversations that are still unresolved, hanging, generating friction nobody can locate because nobody can see it.

Flores called this the network of conversations. And he made a claim that sounds extreme until you sit with it, the health of an organization is identical to the health of its network of conversations. Not correlated with. Identical to.

Fix the conversations and you fix the organization. Because the organization is the conversations.

Every commitment has a structure. Flores mapped it precisely.

A request is made. It has a speaker, the person making the request. A listener, the person receiving it. Conditions of satisfaction, what would count as the request being fulfilled. And a time frame, by when.

If any of those four elements is missing or ambiguous, the request is incomplete. Not just unclear, structurally incomplete. A social fact with a hole in it. And that hole will generate a breakdown later, reliably, in proportion to how important the request was and how ambiguous the missing element.

The listener has four possible responses. Accept, creating a promise. Decline, closing the request cleanly. Counter-offer, proposing different conditions of satisfaction or time frame. Or commit to respond later, keeping the request open while the listener assesses their capacity.

Each response is itself a speech act that either closes the loop or opens a new one. A healthy network of conversations is one where loops close cleanly, where requests meet genuine responses, where promises are kept or renegotiated explicitly before they’re broken, where assessments are spoken rather than accumulated silently into resentment.

An unhealthy network is one where loops don’t close. Where requests are made and acknowledged without genuine acceptance. Where promises are broken without renegotiation. Where assessments are private. Where declarations create realities that different people interpret differently and nobody notices.

This is what Flores means when he says organizational dysfunction is almost always a conversation problem. The project that failed didn’t fail because the technology was wrong or the market shifted, it failed because a request was never clearly made, or a promise was assumed rather than spoken, or conditions of satisfaction were never agreed on, so when delivery happened both parties were surprised to discover they’d been building different things.

The org chart can’t show you any of this. Which is why reorganizing the chart almost never fixes the actual problem.

Now extend this one level further. Flores saw that identity itself, who you are in any context, is constituted by the commitments you’re living inside. Your commitments to others. Their commitments to you. The open requests you’re carrying. The promises you’ve made that aren’t yet fulfilled. The assessments others hold of you based on your history of keeping or breaking commitments.

You are not a fixed entity who then enters into commitments. You are, in significant part, the network of commitments you inhabit.

This is why betrayal feels like more than disappointment. A broken promise doesn’t just fail to deliver, it retroactively alters the reality the promise created. It reaches back in time and changes what was real. The relationship that existed inside the promise is revealed to have been something other than what it appeared. That’s not a small thing. That’s an ontological disruption.

You are currently building commitment networks with AI that can only hold one side of the structure.

You make requests. They’re received and responded to. The conditions of satisfaction are interpreted, sometimes accurately, sometimes not. But there is no promise on the other side. No accountability across time. No identity constituted by a history of kept and broken commitments. No renegotiation possible because there’s nothing to renegotiate with, the next conversation begins with no memory of what was open, what was pending, what was assumed.

Every session starts with a clean network. No loops open. No loops closed. No history.

Which means you can’t build the kind of coordination with AI that you build with people. Not yet. Possibly not ever in the same sense. Because coordination in Flores’ model isn’t just task completion, it’s the accumulation of a shared commitment history that creates trust, predictability, and the background from which genuinely complex joint action becomes possible.

What you can do, and this is genuinely valuable, is use the framework to make your own side of the interaction more complete. Make real requests. State conditions of satisfaction explicitly. Establish time frames. Treat the response as an attempt at fulfillment and assess it against the conditions you actually specified.

You won’t get a real promise back. But you’ll get something much closer to a real response.

And in a world where most people are making gesture-requests and wondering why they get gesture-responses, that’s already a significant advantage.

The network of conversations is the organization. The quality of your requests is the quality of your thinking made visible.

Speak completely. The rest follows.

@ Lee Greene

Leave a Reply

Your email address will not be published. Required fields are marked *