Module 1 — Foundations · Lesson 1.2
The KVN Methodology
Know-How, Vision, Negation — the three axes every delegation must carry
~14 min
What you'll learn
- Define Know-How, Vision, and Negation and tell them apart in practice
- Recognize the failure mode each axis prevents
- Apply KVN at the right scope (workspace, project, task) for the right kind of decision
- Explain why an AI-amplified team needs all three made explicit
Every act of delegation is, in some sense, an attempt to put intention into another head — human or machine. When the intention transfers cleanly, the work comes back recognizable. When it does not, you get drift: well-executed output that solves the wrong problem. KVN names the three axes along which the transfer most often fails, and forces the delegator to be explicit on each.
K — Know-How
Know-How is what the doer needs to be capable of, plus the context they need access to. It answers the question 'do they have what it takes?' before the work begins.
This includes obvious things — relevant skills, prior code, the right credentials, a deployable environment — and less obvious things — the customer's history, the institutional taste, the embarrassing decisions made last quarter that nobody wants to repeat. Know-How is the bag of context and capability the doer carries into the work.
The failure mode Know-How prevents is the silent skill mismatch. You hand a task to someone (or some agent) who plausibly could do it but does not have the specific context the work depends on. They produce something that looks right at a distance and is wrong in a way only the institutional memory would catch. Explicitly naming Know-How — even as a one-line 'needs access to the staging Postgres and familiarity with our auth middleware' — surfaces the gap before it costs you a week.
V — Vision
Vision is the end-state the work is in service of. Not the task description — the outcome. A task says 'implement endpoint X.' A vision says 'so a logged-in customer can finish checkout without leaving the SPA, even on a flaky connection.' The difference is enormous.
Vision is where most management instruction stops too early. The doer hears the task and runs. The doer does not hear the end-state, and so cannot use judgment when the task description is incomplete — which it always is. Without Vision, every small ambiguity becomes a request for clarification or, worse, a guess that resolves in the wrong direction.
Vision is also what lets the AI-amplified worker correct themselves. When an agent encounters a fork — 'should this be a 401 or a 403 here?' — it can answer from the Vision if you have stated one, and from its own priors if you have not. The priors will be plausible. They will sometimes be wrong. The Vision is the only mechanism that keeps the work pointed at your outcome instead of the model's defaults.
N — Negation
Negation is the boundary the work must not cross. The things explicitly out of scope. The trade-offs the doer is not authorized to make. The patterns this team has rejected and does not want to relitigate. Negation is the hardest of the three to write because it requires you to have already learned the lesson at least once.
Every management framework tells you to say what to do. Almost none teach you to say what not to do. The result is doers who, in the absence of explicit nos, fill the gap with their own preferences — often by reaching for the most recently learned pattern, even when the team has already debated and rejected it.
Negation is what makes a delegation safe to leave alone. With Know-How and Vision but no Negation, you have to check in on the work, because you do not know which trade-offs the doer might quietly make on your behalf. With Negation written, the doer (human or AI) can charge ahead at full speed inside the fence and stop cleanly at the edge. Negation is the axis that turns supervision into trust.
Why all three, and why now
Any single axis on its own underdetermines the work. Know-How without Vision produces capable people doing the wrong thing. Vision without Know-How produces ambitious people who cannot execute. Vision without Negation produces ambitious people who execute beyond their authority and ship a regrettable decision. Know-How and Negation without Vision produces well-bounded, capable people doing something that has no point.
The three together compose. Know-How fixes the floor — the work is at least possible. Vision sets the ceiling — the work is in service of something. Negation draws the walls — the work cannot wander outside the room.
Kavanah forces KVN at three scopes because each scope leaks into the next. The workspace charter — your standing K, V, and N for the company — is the default every project inherits. The project charter — K, V, and N for this project — refines the workspace defaults to the project's specifics, including the project's 'false friends' (other projects that look similar and are not). The task K, V, and N — set when needed, generated by AI for routine work — is the per-delegation override. When a question comes up mid-work, you ladder up: check task KVN, then project KVN, then workspace KVN. Almost always the answer is already written down once. The hard part was writing it.
Start your KVN charter
- 1
Open /workspace-kvn and write a one-sentence Know-How
What capabilities does someone — human or AI Employee — need to be effective in this workspace? The first version is rough on purpose. You will refine it once.
- 2
What does the workspace exist to produce? The customer, the outcome, the form of value. Avoid adjectives. Name the thing.
- 3
Write three explicit Negations
Three things this workspace will not pursue, no matter how good the opportunity looks. The point is to have written them once; you can edit later.
- 4
Click 'Generate' on any axis you got stuck on
The AI uses the two axes you filled to propose the third. Treat its draft as a starting point — review, then save.
How to know your KVN is working
- KVN completeness ratio
- Fraction of in-flight projects whose charter has all three axes filled. Kavanah does not roll this up for you yet — read it off the project charters directly (each project's Overview tab), or just keep the count on the same page as your project list. It is a small enough number to eyeball and an important enough one to look at.
- Healthy signal: A Kavanah default, not a measured constant: aim above 90% for committed projects. Lower for exploratory ones is fine — an exploration whose Vision you can already state precisely is usually not an exploration. Lower for committed ones is the red flag.
- Negation reuse
- How often an existing Negation actually gets cited. Kavanah records a hit when the gate refuses work the agent tried to create, when a human dismisses a triage candidate as out-of-scope against a charter line, and when an ingest (a meeting transcript, an import) skips a row that crosses one. Read it on Workspace → KVN, under 'What we said no to' — with the most-cited lines, and how often each was overridden. Note that citing a Negation in a comment or in conversation records nothing; only the paths above count.
- Healthy signal: Rising, or steady with the charter being edited. A Negation with zero hits is either irrelevant or invisible, and those need opposite fixes — so go and find out which before deleting it. A Negation with a high override count is the loudest signal here: it is a rule the team has quietly stopped believing in, and it wants a decision rather than another override.
- Clarification rate
- Fraction of tasks that need a clarifying comment before the doer can start — a proxy for incomplete KVN. Kavanah does not compute this; it is a hand count over a week of tasks, and worth doing once a quarter rather than never.
- Healthy signal: A Kavanah default, not a measured constant: under roughly 10%. If it runs higher, your charters are missing the axes the work hits first — and the comments themselves tell you which axis, because people ask about what was not written down.
Key takeaways
- ·K is the bag of context and capability the doer carries in.
- ·V is the end-state the work is in service of, not the task description.
- ·N is the boundary the work must not cross — the hardest to write and the most load-bearing in an AI-amplified team.
- ·All three together compose; any one alone underdetermines the work.
- ·Kavanah enforces KVN at workspace, project, and task scope so the answer is usually already written down once.
KVN is the spine of everything else in this course. The Negation is the axis the product actually enforces: it is checked before any task or project the agent creates, and work that crosses it is refused rather than filed. Vision sharpens that judgement — it is what lets the check tell a false friend from real work — but on its own it enforces nothing, so a charter with a Vision and no Negation is the most misleading state you can be in. Know-How is the axis you write for the doer, human or AI: today it is context they read, not a signal the assignment router scores on, so treat it as instruction rather than automation. From here on, every lesson assumes you have a draft charter to refine, not a blank page.
Sources
- 1.What Is Strategy?
Michael E. Porter · Harvard Business Review 74(6) · 1996
'Trade-offs are essential to strategy… they purposefully limit what a company offers' — the canonical statement that strategy is defined by what you deliberately do not do, the basis for the Negation.
- 2.Building a Practically Useful Theory of Goal Setting and Task Motivation: A 35-Year Odyssey
Edwin A. Locke, Gary P. Latham · American Psychologist 57(9), 705–717 · 2002
Specific, difficult goals reliably outperform vague 'do your best' instructions — why stating the Vision beats leaving intent implicit.
- 3.The Effects of Feedback Interventions on Performance: A Historical Review, a Meta-Analysis, and a Preliminary Feedback Intervention Theory
Avraham N. Kluger, Angelo DeNisi · Psychological Bulletin 119(2), 254–284 · 1996
Meta-analysis of 607 effect sizes: over a third of feedback interventions decreased performance — supports that badly-specified guidance resolves ambiguity in the wrong direction.