Module 1 — Foundations · Lesson 1.3
The Four Primitives
Direct, allocate, feedback, consequence — and what each looks like in Kavanah
~11 min
What you'll learn
- Recognize each primitive in real situations
- Diagnose which primitive is failing when a project is stuck
- Map each primitive to the Kavanah surface that supports it
- Avoid the most common over-investment trap for each
Frameworks are choreography. Primitives are physics. The four primitives — direction, allocation, feedback, consequence — recur in every era of management because they are what coordination is. Strip away the rituals and what you have left is some choice about each of the four, made well or made poorly. This lesson walks each in turn, names the failure mode that follows when it is starved, and shows you where in Kavanah it lives.
Direction: naming the outcome
Direction is the answer to 'what are we producing?' It is the V in KVN at workspace scope, the project charter at project scope, and the task description at task scope. It seems obvious until you discover, in a stand-up, that three of your six people would phrase the project's outcome differently if asked separately. At that point it is not direction; it is a shared assumption that has not been tested.
The failure mode of starving direction is parallel rationality — every contributor making sensible local decisions that do not add up to anything coherent. The work product reads like several people writing chapters of a book they have not agreed on.
The over-investment trap is the opposite: spending so long perfecting the direction that you never get to test it against reality. The best test of a stated direction is not whether the leadership team agrees on it; it is whether a junior contributor can recite it without looking, three weeks after it was written.
In Kavanah, direction lives in the workspace and project KVN charters, and is surfaced into every AI agent conversation as system context. The AI agent's job is, in part, to remind the team of stated direction when proposed work drifts from it.
Allocation: matching attention to direction
Allocation is the answer to 'who is doing what?' It is the slowest of the four to feel painful and the most expensive when wrong. Direction is cheap to change; allocation is expensive, because once a person has been pointed at a problem they accumulate context that does not transfer.
The failure mode of starving allocation is silent thrash — the team is busy, but the work that ships does not move the needle, because the most senior or most capable people are spending their time on the most urgent rather than the most important.
The over-investment trap is allocation theater: weekly re-planning meetings that move tickets between columns without changing who is actually doing what next Monday. Allocation has done its job when, on any given day, every contributor can name the one thing they are working on and can explain why it is the right thing.
Kavanah's allocation surfaces are the Projects board, the Capacity and Availability views, the skills-based assignment recommendations, and the AI agent's ability to recommend assignees from a task description. Module 3 covers each in detail.
Feedback: knowing whether it is working
Feedback is the answer to 'are we on track?' Specifically — at a fine enough grain that course-correction is still cheap. A retrospective is feedback; it is just feedback delivered too late to change the project being retro'd. Useful feedback is the kind you get while the work is in flight and can still be redirected.
The failure mode of starving feedback is the late surprise: you find out a project is in trouble on its due date, by which point the only available options are the most expensive ones. Most of management's job, on a quiet week, is engineering feedback loops short enough that the next surprise is not a late one.
The over-investment trap is feedback theater: dashboards no one looks at, daily standups where everyone reports green, status reports written for the audience above and not for the team itself. The test of a feedback loop is not its existence — it is whether someone visibly changed their plan because of it last week.
In Kavanah, feedback is the daily AI agent review, the dashboard cards, the Insights → Reports surface, the build-status and observability cards on /dev, and — most quietly — the comment trail on each task. The metric framework in Module 6 is largely a framework for engineering feedback loops you will actually read.
Consequence: what happens after
Consequence is the answer to 'did it matter?' It is the loop that turns feedback into changed behavior. Without consequence, the other three primitives become theater within one or two cycles, because the team learns that nothing they say or do actually changes what gets prioritized, who gets promoted, or what work gets continued.
Consequence does not have to mean punishment. The cleanest forms of consequence are positive: the project that worked gets the next investment, the person who shipped well gets the next high-leverage assignment, the practice that produced the win gets adopted by the next team. The point is that the outcome of last quarter visibly shapes the inputs to this quarter.
The failure mode of starving consequence is what most large organizations exhibit: feedback is collected, retrospectives are run, lessons are 'captured,' and next quarter looks identical to last quarter. The lessons did not bind. There was no consequence attached.
The over-investment trap is performative consequence — public reorgs, dramatic firings, sweeping policy changes — that signal change without producing it. Real consequence is small, frequent, and proportionate. Kavanah supports consequence through the skills-reinforcement loop (the people who shipped the work get credited with the skill, which then routes them more of it), the retrospective surfaces in Knowledge, and the project portfolio's history of which charters got renewed and which got killed.
Diagnosing which primitive is failing
When a team is stuck, the fix is almost always one primitive, not all four. The diagnostic is usually fast:
If the team is busy but the work product is incoherent — direction is starved.
If the team is busy but the most important problem is not getting attention — allocation is starved.
If the team is moving but you keep being surprised by late bad news — feedback is starved.
If the team is moving but the same mistake keeps recurring quarter to quarter — consequence is starved.
The wrong move, every time, is to try to fix the symptom with more of a different primitive. More feedback will not fix a direction problem. More allocation will not fix a consequence problem. Each primitive responds to its own intervention.
Walk your workspace through the four
- 1
Direction — open /workspace-kvn
Confirm the workspace and one active project have a written Vision. If a project does not, this is your first homework.
- 2
Check the Resources view. Does every active project have a named lead? Does every member have at most one 'most important' task in flight?
- 3
Feedback — open the Dashboard's AI agent review
Read the agent's current summary of work in flight. If it surfaces a surprise, the feedback loop is doing its job.
- 4
Consequence — open /knowledge → retrospectives
Pull up the most recent retro. Are any of its lessons actually visible in this quarter's plan? If not, the lesson did not bind.
One metric per primitive
- Direction recall
- Fraction of team members who, asked separately, recite the same one-sentence project outcome.
- Healthy signal: Above 80%. Below 50% means your charter is on paper, not in heads.
- Allocation focus
- Median number of 'in progress' tasks per person across the workspace.
- Healthy signal: 1.0 to 2.0. Above 3 means allocation is not being made; it is being defaulted.
- Feedback latency
- Time between a project going off-track and a stakeholder finding out.
- Healthy signal: Hours to one day. Weeks is the late-surprise zone.
- Lesson bind rate
- Fraction of explicit retrospective decisions that produce a visible change in the next plan.
- Healthy signal: Above 60%. Lower means retros are documentation, not consequence.
Key takeaways
- ·The four primitives — direction, allocation, feedback, consequence — are what every framework is choreographing.
- ·Each primitive has a distinct failure mode, an over-investment trap, and a Kavanah surface.
- ·The wrong move is to fix one primitive with more of a different one.
- ·Diagnose what is starved; intervene there.
From here, the course walks Kavanah's surfaces in roughly the order the primitives demand attention. We start with what direction looks like in the wild — and where work is being lost before it ever gets named.