Kognit Take the test
Framework · Version 1

The Four Behavioral Axes of Working with AI

Two people, same job, same tool — and completely different work. One slices every task into bite-sized pieces and double-checks every number; the other ships whole work packages to the model and only reads the result. The difference isn’t skill, and it isn’t the model. It’s behavior — and it can be described along four axes.

~7 minute readThe model behind arbeitstyp.deLast updated: July 2026
The wrong question

“How much do you use AI?” measures the wrong thing.

Most surveys and team retros ask the intensity question: daily, weekly, never? That’s about as revealing as asking “How much do you use the phone?” in 1995. What’s interesting isn’t how often someone opens a language model — it’s what shape the collaboration takes.

Someone who uses AI daily can still treat it like a better search engine: question in, answer out, back to the old routine. And someone who opens it twice a week might be delegating entire work packages with a clear brief and acceptance criteria. Same “usage intensity,” two entirely different ways of working. A useful model therefore has to describe behavior, not frequency. Four questions are enough to span the shape of the collaboration.

The model

Four axes that actually discriminate.

Axis 1 · Delegation

Assistant ↔ Autopilot

How much do you hand over? At one pole, the AI gets small, tightly scoped subtasks — the next paragraph, that one formula. At the other, it gets whole tasks with context and delivers results, not building blocks.

Axis 2 · Verification

Verifier ↔ Truster

How much do you verify? Verifiers check sources, numbers, and logic before anything moves on. Trusters take convincing output at face value — faster, but with a different risk profile.

Axis 3 · Focus

Thinking partner ↔ Maker

What do you use AI for? As a sparring partner for thinking — sharpening ideas, testing structures, hearing counterarguments? Or as a production machine for finished output — text, spreadsheets, code?

Axis 4 · Method

Architect ↔ Improviser

How do you steer? Architects build reusable systems: prompt templates, fixed routines, building blocks. Improvisers feel their way forward in open dialogue and let the prompt emerge from the conversation.

Important: neither side is “better.” A Verifier drafting legal briefs is exactly where they should be; so is a Truster generating brainstorming material. Things only go wrong when axis position and task don’t match — the Truster in contract law, the Verifier spending three hours fact-checking a mood board. Making that fit visible is what the model is for.

Methodology

How the test measures — and where its limits are.

The Kognit test operationalizes each axis with six statements, 24 in total, answered on a five-point agreement scale. Three design decisions are worth noting:

1

Mixed polarity

Half the statements are worded in favor of one pole, half in favor of the other (“I prefer handing the AI a whole task” vs. “I prefer breaking tasks into small pieces”). Agree with everything and you land in the middle, not at the edge — acquiescence bias gets dampened instead of rewarded.

2

Behavior, not attitude

The statements describe concrete work behavior (“I routinely fact-check what the AI claims”), not opinions about AI (“AI will transform the workplace”). Behavior sits closer to what actually happens day to day — and is easier to answer honestly.

3

Deterministic scoring

The four axis scores map to one of 16 types — transparent and reproducible, no black box. Answer the same questionnaire the same way and you get the same type.

And the limits, no hedging: the model is theory-driven in its construction but not yet validated on a large sample — reliability and structural checks are pending and will follow once enough real response data exists. It is self-report, with all the known biases. Until then, the role benchmarks inside the test are explicitly labeled as a modeled preview, not a measurement. We publish no prevalence claims (“X% of consultants are type Y”) before there is solid data.

The typology

16 types from four letters.

Each combination of axis positions yields a type with its own four-letter code — from EPDA, the Watchmaker, to WVMI, the Pioneer. (The letters come from the original German axis poles; treat them as symbols, like any four-letter typology.) The names are mnemonics, not boxes: the code behind them shows your four positions, and those are the actual result.

E P D A

The Watchmaker

“Precision over speed — every gear in its place.”

E P D I

The Socratic

“Thinks it through — in dialogue, not alone.”

E P M A

The Craftsman

“Craftsmanship with AI, controlled and repeatable.”

E P M I

The Tinkerer

“Hands-on, testing, always fiddling.”

E V D A

The Strategist

“Small moves, big plan, trust in the tool.”

E V D I

The Idea Surfer

“Plays, leaps, trusts the spark.”

E V M A

The Curator

“Selects with taste — and trusts the selection.”

E V M I

The Sprinter

“From idea to result, no detours.”

W P D A

The Commander

“Delegates at scale, checks where it counts.”

W P D I

The Researcher

“Hands off a lot, questions everything, follows the trail.”

W P M A

The Producer

“Builds pipelines and inspects the output gate.”

W P M I

The Generalist

“High volume, high speed, verified — whatever comes in.”

W V D A

The Visionary

“Big picture, big trust, clear frame.”

W V D I

The Alchemist

“Mixes, experiments, trusts the reaction.”

W V M A

The Conductor

“Orchestrates systems and lets them run.”

W V M I

The Pioneer

“Maximum delegation, maximum speed, always ahead.”

So which type are you?

24 questions, about 4 minutes, no sign-up. Your archetype, your axis profile, and your blind spots — free.

Take the test — free
The full report (prompt library, delegation map, workflows) is a one-time €14.