What Management Actually Is, Now

The four primitives every era rediscovers

Across cuneiform ration tablets, legion rolls, monastic horaria, factory time-and-motion studies, and modern OKR documents, four operations recur. Call them the primitives.

Direction. Naming the outcome the group is trying to produce. Without it, work fragments into individually rational but collectively pointless effort.

Allocation. Matching people, time, money, and attention to that outcome. The hardest of the four because every yes is also a no.

Feedback. Knowing whether the work is hitting the outcome — fast enough to course-correct, not learn at the post-mortem.

Consequence. Rewarding what worked, correcting what did not, removing what cannot be corrected. Without consequence, the other three decay into theater.

Every framework you have ever heard of — Sun Tzu, Drucker, Deming, Six Sigma, Agile, OKRs, Shape Up — is a different choreography of these four. The frameworks differ. The primitives do not.

What changed in the last three years

For most of the discipline's history, the binding constraint on output was human throughput. A team could only ship as much as it could write, design, or build by hand. Management was, in large part, a craft of squeezing more work out of a fixed labor pool — through scheduling, motivation, division of labor, and the steady removal of friction.

Generative AI broke that constraint — on the right work. Controlled trials find large gains on well-bounded tasks: writing time down about 40% with quality up (Noy and Zhang 2023), a coding task finished 56% faster with an AI assistant (Peng et al. 2023), consultants inside the AI's competence 25% faster at higher quality (Dell'Acqua et al. 2023). On the right task, a single contributor with a competent agent can produce in a day what used to take a team a week. But the gain is uneven, not a uniform multiplier: in a 2025 randomized trial, experienced developers working in their own mature codebases were actually 19% slower with AI while believing they were faster (METR 2025), and across a large industry survey a jump in AI adoption came with a small drop in delivery throughput and stability (DORA 2024). The bottleneck still moved. It is no longer how much the team can produce — it is whether what they produce is the right thing, in the right order, at a quality bar a customer will pay for.

That shift moves the manager's job decisively up the four primitives. Direction matters more, because on the work AI does accelerate, the same labor hour can create far more output — and far more drift if pointed at the wrong outcome. Allocation matters more, because attention is the only scarce resource left. Feedback matters more, because the gap between 'we shipped it' and 'it was wrong' now takes hours instead of weeks. Consequence still matters exactly as much as it ever did, which is to say it is the rate-limiting step on whether anything you learn actually changes behavior.

Alignment, not throughput, is the new bottleneck

The cheap mental model is to imagine AI as a faster pair of hands. The accurate mental model is to imagine AI as a junior employee who will faithfully execute the ambiguity in your instructions, including the ambiguity you did not realize was there. Hand it a task with only positive direction — 'build this' — and it will fill the unspoken space with its own priors, not yours. The output looks plausible. It is also, often, not what you actually wanted.

The practical consequence is that the parts of management most people used to consider 'soft' — clarifying intent, naming what is in scope, naming what is explicitly out of scope, deciding what 'good' looks like — have become hard load-bearing requirements. You can no longer compensate for vague direction by personally re-doing the work. The team will keep going at AI speed, in whichever direction the ambiguity pushes them, and you will discover the misalignment days or weeks later.

Kavanah is built around this premise. The KVN methodology — Know-How, Vision, Negation — is an explicit forcing function for the three axes where ambiguity hides. The whole product, from the way conversations become tasks to the way personas are scoped, is an attempt to make alignment cheap enough to do every day.

What this course is about

This is a project management course written for the moment we are in. It teaches the discipline as it has always been practiced — the four primitives, the cadence, the metrics — and grounds every part of it in the specific surfaces Kavanah provides. Each lesson is short. Each ends with a concrete action you take inside the product, and a metric you watch to see whether the action is working. By the end, you will have stood up a workspace, captured your team's standing intentions, instrumented a measurable operating rhythm, and routed the right work to the right people — some of whom may be AI Employees.

The course assumes no prior management experience and no prior Kavanah experience. It does assume you are interested in why each technique exists, not just how to click it. If you skip the why, the techniques become rituals — and rituals decay the moment the team you learned them on changes.