Four practical questions for analysing your own performance when working with AI

A descriptive practitioner note on holistic performance analysis applied to human-AI collaboration and AI-intensive knowledge work. It passes one's own working process through four sequential questions: is it observable, can it be decomposed, can it be measured honestly, and which part is within my own control to change? The framework is designed for knowledge workers operating with AI tools in complex, multi-step or multi-system environments — where the human-side bottleneck is regularly misattributed to the AI model. The four-lens approach applies wherever a human operator works with AI systems in demanding environments: knowledge engineering and ontology development, multi-agent AI architectures, autonomous systems and robotics, swarm-based AI coordination, and AI-augmented research or decision workflows. The building blocks (improvement loops, bottleneck thinking, skill decomposition, situation awareness in complex systems) are established and explicitly credited to prior work (Shewhart/Deming, Boyd, Goldratt, Ericsson 1993, Endsley 1995). The specific contribution is the transfer to human-AI collaboration as the unit of analysis: a forced human-vs-tool factor split, built-in measurement honesty (each indicator carries an explicit note of what it does not measure), and focus on controllable human-side factors in AI-augmented workflows. Three predictions are stated as open, untested hypotheses (H1-H3). This is a descriptive framework — not a test, scoring, or ranking instrument; not legal or medical guidance; not empirically validated. Origin: transfer from high-performance coaching methodology. Includes a German Zusammenfassung. License: CC BY-NC-ND 4.0.