Confessions on a Dance Floor as a mental model for Agent Experience Optimization.
There was a precise moment when Madonna stopped making pop “for radio” and started speaking directly to DJs.
That moment was Confessions on a Dance Floor, 2005: an album conceived as a single continuous set, with tracks built around mixable intros and outros, structures designed to be used, not just heard. Before that album, the standard path was: single comes out, then remixes arrive for the DJs, then — maybe — the track makes it into the night. With Confessions, Madonna inverts the order: she designs the album as material to be mixed from the start.
Something analogous is happening with digital presences and AI agents. But to understand what, you first need to understand what a DJ actually does.
The DJ doesn’t consume and doesn’t narrate: they act
A DJ isn’t a listener with headphones, and isn’t a remixer creating new tracks. Their goal is one: build a set that makes people dance. They have no authorial claim on the material they use, no explanatory or narrative ambition. They take different tracks, break them down into their functional components — the intro, the loop, the outro — combine them in sequence, and produce something that wasn’t in any of the source material: movement on the floor.
An AI agent works exactly like this. It doesn’t read your digital presence to understand it or cite it. It moves through it to act: it looks for what it can do, combines that with what it can do elsewhere, and builds a process that leads to a result. It books, compares, selects, purchases, fills in, returns. The goal is purely active — and this changes completely what a digital presence needs to be in order to function in the age of agents.
The problem isn’t content: it’s the structure of actions
When we say a digital presence needs to be “mixable” for an agent, we’re not talking about well-written articles or text optimized to be cited. We’re talking about endpoints, available actions, processes that an agent can move through autonomously. An e-commerce shop, a services site, a platform: any digital property, from an agent’s point of view, is a set of things that can — or cannot — be done.
A “pure pop” track can get away with irregular structures, unpredictable entries, abrupt transitions. For a listener, that’s fine. For a DJ building a set, it’s a problem: they don’t know where to come in, don’t know where to get out, don’t know how to connect it to what comes before and after. The track stays out of the set — not because it’s bad, but because it wasn’t built to be used by someone else with their own goal.
Today’s digital presences are almost entirely designed for the human listener: structures built for navigation, for scrolling, for perceptual experience. Rarely are they designed for an actor that doesn’t browse, doesn’t scroll, isn’t drawn in by a homepage — but enters, looks for what it can do, does it, and exits with a result.
Intros and outros: what they actually mean in a digital presence
If the metaphor holds — and it does — then intros and outros have a precise meaning that goes well beyond textual structure.
The intro of a digital presence, for an agent, is the map of what can be done there: which actions are available, through which tools, under which constraints. It’s not an “About” page, nor a service description written to convince a human. It’s the answer to the question an agent asks before anything else: what can I do here? If that answer isn’t exposed in a stable, readable way — if the available actions are buried in dynamic structures, in flows designed for human navigation, in pages built to seduce rather than to enable — the agent doesn’t enter the set. It moves to another property.
The outro is what the agent returns to the user after acting: the final state of the process, the concrete result, what they take away from the interaction. In an e-commerce context, it’s the purchase confirmation with all verified data. In a service context, it’s the appointment booked, the form submitted, the recommendation produced. The outro isn’t a thank-you: it’s proof that the process was completed reliably and that the user got what they delegated.
Between intro and outro is the loop — the internal structure of actions, the steps that need to be deterministic, verifiable, unambiguous. This is where most of the failure points we see today are concentrated: availability not surfaced, data accessible only via JavaScript, overlays intercepting actions, flows that change state without signaling it.
Why this is AXO and not an optimization question
SEO asked: can they find me? GEO asked: do they cite me? AXO asks something different: can they use me? Can they enter, act, and exit with a result?
It’s not one more step on the same scale. It’s a change of interlocutor. The DJ doesn’t need the track to be beautiful — they need it to be mixable. A beautiful track without a clean intro, that changes BPM without warning, that has no outro to lean on for the transition, stays out of the set. Not because the DJ doesn’t appreciate it: because they can’t use it.
Designing for actionability means accepting that a growing share of interactions with your digital presence will no longer pass through a human who browses and decides — they’ll pass through an agent that enters with an objective, moves through the system, and returns a result to someone who never opened your site. If that traversal isn’t possible, if the system wasn’t built to be moved through autonomously, your property isn’t poorly optimized for that ecosystem: it simply doesn’t exist.
If you want to understand whether your digital presence is already actionable or needs to be redesigned, write to me.