The Sun King in the Land of AI Agents

The upheaval is brutal. Until yesterday, all we had to move the needle were our own modest efforts and our limited time. Today, our capacity to produce no longer stops at what we can carry out ourselves. A single person can mobilize a workforce that would once have required an entire team. And yet most businesses still approach work with the reflexes of the old world. Playing the same strategy makes no sense when your means have changed this much.

I chose to turn to the figure of Louis XIV. The Sun King designed a relentless system of command, and that system built Versailles.

Work. A great deal of work. The Sun King defined sovereignty as a demanding daily labor. He worked three times a day on a rigorous schedule: councils in the morning, finance included, the routine business of the state in the afternoon, and in the evening, alone in his study or with Louvois for the details of war. Absolute power took the form of effort. He also kept notes and dictated a journal, which his secretaries would later shape into his Mémoires. In them he judged his own actions, and in every undertaking he looked for a chance to establish his greatness.

That attention can settle on a tiny detail if the detail decides the overall effect. It does not require every micro-decision to come back to us. It requires us to know where our judgment counts. The machine can multiply proposals. It does not excuse us from training our eye.

The Sun King was obsessed with the effect Versailles had to produce. He went as far as writing a guide himself, Manière de montrer les jardins de Versailles (the way to present the gardens of Versailles), which describes 25 precise steps for touring the grounds and went through 7 successive versions between 1689 and 1705, some of them in his own hand! His guide imposes a choreography of the gaze on the visitor: where to stop, where to turn, what to consider, with every detail meant to display royal power.

Versailles invites a question: what should the person who enters here feel? That question gives direction to the thousands of decisions that follow. For our own productions, the effect we seek might be trust, surprise, clarity, or pleasure. Still, we have to choose it. Without that intention, abundant means mostly let us pile up elements that add up to nothing. At Versailles, intention takes the form of spectacle. In our own works, it can just as well express itself through radical simplicity. It is the strength of intention, asserted down to the details, that lifts us out of the ocean of commodities generated by algorithms.

In 1684, as gilders worked night and day in the king’s new bedchamber to be ready for the monarch’s arrival, he complained about the foul smell of the oily gilding size (or-couleur). Louvois and the gilder devised a machine to heat the room, and Bontemps, the steward of Versailles, had fires lit and the room aired to clear the stench before the king retired for the night.

This is where caprice becomes fertile. “I want it anyway” pushes us to search beyond the first acceptable solution. You can send everything back for a nuance, a sensation, an extravagant idea you cannot yet fully explain. As long as we learn from every iteration, that stubbornness shows us how far we can really go.

To work together at a distance under Louvois (1683-1691), the king and his superintendent used sheets of paper folded in half lengthwise. The right-hand column was reserved for the minister’s text, while the left half was left entirely blank so the sovereign could write his replies by hand.

A simple “BON” from the king, French for “approved,” was enough to launch a major project. That “BON” was only the visible part of work already done: a matter investigated, a proposal drafted, a decision made possible. The brevity of the answer rests on the quality of the preparation. If every approval forces us to rebuild the whole problem, we have only moved the work around.

Today, our AIs have to step into the role of those devoted ministers, synthesizing the data and structuring the options to prepare our decisions. The decisions that need our arbitration must reach us in a clear form: a recommendation, its reasons, and its consequences. A simple “BON” can then be enough to set things in motion. Routine choices stay delegated, within a framework defined in advance.

The Palace of Versailles is the work of a dizzying number of contributors, filling all kinds of roles and succeeding one another over time. To make such a work possible, the royal administration developed a relentless engineering of information. Every order had to leave a written trace. The art of recording everything is tedious, but its benefits are concrete: on the building site of the Grande Écurie, it was the methodical comparison of the theoretical measurements recorded in the ledgers with the bills submitted by the contractors that produced material proof of embezzlement.

The system would outlive the king: under Louis XV, the King’s Buildings office even set up a “bureau des discussions” to filter and record every request. Delegating massively without losing control requires information that is kept up to date and within reach. With agents, we call it company context: projects and their history, decisions and their reasons, budget, schedule, and regulatory constraints, clients, know-how written down as skills. An agent with access to it works like a well-briefed colleague. Without that context, it starts from scratch with every request, and we are the ones filling the gaps. In the organizations I work with, this information almost always exists, but it is scattered across people’s memories, email threads, and files no one can find. Bringing it together is the first job, long before building an AI agent.

Every correction should leave behind more than a better deliverable: reusable knowledge of what we are looking for. Why was this proposal rejected? What made the next one right? Which choices are now settled? Recording these answers lets the next projects start from higher ground. Without that memory, we are condemned to explain our ambition all over again and to win back the same details.