Planning at Altitude
You have been promoted, whether you asked or not. The moment AI could write the code, draft the copy, and run the research, everyone became the executive of everything they run. Almost nobody was given the tools for the new job.
There is a two-century-old answer to leading work you cannot personally do, and it comes from the least fashionable place: military command. Prussian officers called it Auftragstaktik, mission command. The commander states the intent, the outcome that must be true and why, then delegates the method entirely to the people on the ground, who understand their terrain better than any headquarters can. The commander's job collapses into three things: make the intent unmistakably clear, hold the map, and resist the temptation to reach down and micromanage the method.
Every part of that translates to working with AI agents, and the failure modes translate too. Agents are tireless, fast, and surprisingly capable, and they will happily execute a vague intent into confident nonsense. The research on human-AI teams keeps finding the same bottleneck: not the model's ability, but the human's capacity to specify what they actually want and to supervise ten threads without drowning. Approval fatigue is real: people rubber stamp what they no longer have the bandwidth to genuinely review. The scarce resource is oriented judgment.
The map on the table
Notice what sits at the center of every command post in history: a map. Not a list of the army's tasks, a map: the whole theatre visible at once, positions current, the commander physically above the detail. The map is what makes intent-level thinking possible, because you cannot think at altitude while reading a queue. A queue hands you items; a map hands you the situation.
That is the shape modern work is missing. Your projects have plenty of queues: tickets, inboxes, chats, each one pulling you down to the item level, each context switch costing you the picture. What is missing is the table with the map on it: goals and milestones laid out in space, progress rolling up from wherever work actually happens, so that a glance answers "where are we, really?" and the next decision gets made at the level where you are irreplaceable.
Altitude is a discipline, and tools either support it or erode it
Staying at altitude takes structural support, because gravity pulls the other way: detail is comfortable, legible, and endlessly available. A tool supports altitude when the high-level view is the default and the detail is one deliberate step down, when delegation has a first-class surface rather than a copy-paste ritual, and when progress reports itself instead of being reconstructed in meetings.
Orient is built as that table. The map holds your goals in space; every node keeps score; branches roll their progress up so the root answers for the whole; and agents work the same map through the API that you work on the canvas, under the same permissions. You state the intent and watch the theatre. The method, increasingly, is machine territory, exactly as the old commanders would have wanted it: they never did the marching either.