Skip to content
AI Engineer Guide.

concept · career · advanced · concept 7 of 33

Leveling

0 of 33 complete

Why it matters

Leveling exists to make expectations legible: how much ambiguity someone should absorb, how far their decisions reach and what kind of review they need. Candidates often prepare by listing years and tools, which says nothing about scope. The useful preparation is to identify the decisions you made without escalation, the people who acted on your judgement, and the problems you framed rather than solved. That is the evidence a level conversation is actually about.

Mental model

A level is a radius of responsibility: how far your decisions travel and how much uncertainty you are expected to resolve alone. The metaphor breaks because a radius is visible, while scope is inferred from stories, so you have to make it explicit.

How it works

Scope and reach are different axesInteractive
highReach of the impactlow
one taska system, up to org direction
Scope of the decisions you own →
VerdictChoose· recommended hereConditional· with the named costStandard· the ordinary case

Step 1 of 4

Scope, not tenure

A level is a radius of responsibility, not years served. The evidence starts with a decision you made without approval and what depended on it.

Scope of the decisions you own: low · Reach of the impact: low

Standard · Show the decision you owned

Describe the decision inside the task; years and tools say nothing about scope.

A monthly note of the decisions you made without escalation is the raw material. For each, record what depended on it.

Axes

Scope of the decisions you own
one task → a system, up to org direction
Reach of the impact
your team executes → other teams act on it
Step 1/4 · Scope, not tenure

Figure. A level is a radius of responsibility: what you decide without review, and who acts on it. A narrow decision other teams reuse is not the same evidence as a broad decision your own team executes.

What to notice. The cells map evidence, not titles. Levels are described differently across companies, so no level name or number appears; the top-right cell is the usual promotion case, not a requirement that every level occupy it. The right column holds both a system you shaped and a direction you framed; the axis is ordinal, not a scale. The figure does not claim that one quadrant is better work — it claims that a level conversation asks for the decision behind the work.

Text version

A two-by-two map of level evidence. The horizontal axis is the scope of decisions you own, from one task to a system, up to org direction; it is ordinal, so the right column holds both a system you shaped and a direction you framed. The vertical axis is the reach of the impact, from your team executes at the bottom to other teams act on it at the top. Bottom left, a defined task with the decision inside it: show the decision you owned, because years and tools say nothing about scope. Bottom right, a system you shaped that your own team uses: show the system you shaped, because depth is real scope but narrows influence to the people who depend on it. Top left, a narrow decision other teams act on: show the judgement others reuse, reach without breadth. Top right, a problem you framed and a direction that outlived the project: show the problem you framed, the clearest signal, a decision nobody above you approved. Four steps walk the cells in that order.

  1. Write down the last three decisions you made that nobody above you approved, and what depended on them.
  2. For each, note who acted on it: a team, another team, or a direction that outlived the project.
  3. Identify the problems you framed, not just the ones you were assigned, since framing is the clearest scope signal.
  4. Collect evidence of influence through writing, review or standards rather than through authority.
  5. Compare your scope against the next level's description and find the gap, rather than assuming tenure closes it.
  6. Ask your manager which specific evidence is missing, and turn the answer into one project rather than a general goal.

Trade-offs

ChoiceGood forCostFails when
Optimise for scopeStrong promotion evidenceRisk of unglamorous work being neglectedThe team needs reliability more than expansion
Optimise for depthRecognised expertiseNarrow influence and dependency on youThe organisation needs direction, not a specialist
Stay generalistBroad contribution and mobilityDiffuse evidence at review timeThe level expects a clear area of ownership
Chase the titleImmediate external signalScope may not follow the labelThe next role interviews on scope you have not had
Trade-off mapInteractive

Good for

Strong promotion evidence

Cost

Risk of unglamorous work being neglected

Fails when

The team needs reliability more than expansion

  1. The team needs reliability more than expansion
  2. The organisation needs direction, not a specialist
  3. The level expects a clear area of ownership
  4. The next role interviews on scope you have not had

Figure. Each choice with what it buys, what it costs, and the case where it loses.

What to notice. The fourth column is the decision: if the losing case describes your situation, the choice is wrong.

Text version

Four trade-offs: Optimise for scope is good for Strong promotion evidence, costs Risk of unglamorous work being neglected, and fails when The team needs reliability more than expansion. Optimise for depth is good for Recognised expertise, costs Narrow influence and dependency on you, and fails when The organisation needs direction, not a specialist. Stay generalist is good for Broad contribution and mobility, costs Diffuse evidence at review time, and fails when The level expects a clear area of ownership. Chase the title is good for Immediate external signal, costs Scope may not follow the label, and fails when The next role interviews on scope you have not had

Failure modes

  • Tenure argument. The case rests on time in role and the panel asks for scope evidence. Debug: list decisions from the last year that required no approval and check whether any exist.
  • Technology list. The answer names systems but not consequences. Debug: for each technology, write the decision it enabled and who depended on it.
  • Borrowed impact. Team outcomes are described as personal without a clear contribution. Debug: separate what you decided from what you executed for someone else.

In production

Keep a monthly note of decisions, their reach and their outcome. Review it before any level conversation so the evidence is contemporaneous rather than reconstructed.

Common misconceptions

People think: Levels are comparable across companies.

Actually: They describe different scopes and expectations; the label alone tells you little without the responsibilities attached.

People think: Promotion is a reward for good work.

Actually: It is a prediction that you already operate at the next scope, which is why evidence of that scope is required in advance.

Interview angle

Frame level as scope: decisions owned, ambiguity absorbed, influence beyond the team. Give one concrete decision per claim. Weak answers count years and technologies, or describe a team outcome without separating their contribution.

Sources