Designing AI that makes people perform better
AI should do more than produce answers for people to approve. It should help them think, learn, and make better decisions. The Playbook offers practical guidance for designing this kind of human–AI collaboration, oriented by one grand challenge: build systems where the human learns more from the interaction than the system does.
Hybrid Intelligence in practice
HI Differential Diagnosis
Developed with Systematic, this diagnostic prototype explores how AI can support emergency-department clinicians while keeping human reasoning and judgment central.
From offboarding knowledge to smarter onboarding
Developed with Xboarding.ai, this prototype captures departing employees’ knowledge and turns it into structured onboarding through reports, AI assistance, and reflective quizzes.
Co-create complex offers with AI
Developed with Knowit, this guided workflow uses AI to analyse requirements and rank options while people calibrate the reasoning and make the final decisions.
Turn team signals into human-led action
Developed with Involved, this five-stage workflow helps leaders interpret employee signals, prioritise risks, and act while requiring human judgment at every critical step.
Sleeper Sofa Guide
Developed with Innovation Living, this prototype explores how an AI guide can match customers with the right sleeper sofa while keeping staff in charge of reviewing, tuning and improving its advice.
What we have built internally for our research center
Quantum Career Chatbot
A chatbot that helps users explore career paths, challenge assumptions, and refine options without taking over the decision.
Quantum Strategy Chatbot
A co-pilot for comparing quantum strategies, exploring sources, and viewing policy problems from multiple perspectives.
Abstract
Human-centered AI research has produced deep knowledge about designing interaction so that people reflect, learn, and reach better joint outcomes with machines, yet little of it reaches deployed systems, where generative and agentic tools reposition users from producing work to evaluating machine output; in one landmark 2025 study, physicians assisted by a diagnostic language model performed worse than the model alone, though the assistance carried none of the field’s known scaffolding.
We hypothesize that the binding constraint is operationalization rather than further foundational insight, and report first results of the Hybrid Intelligence (HI) Design Playbook, a community-co-created pattern language oriented by one grand challenge: the human should learn more from the interaction than the system does. Eighty-six researchers contributed 26 patterns, consolidated with documented provenance into; a one-day sprint with 38 practitioners from 22 organizations yielded 13 prototypes, and Playbook-trained students embedded in a manufacturing SME built a parallel automation case.
Twin flagships anchor the results at opposite poles of the complexity landscape: a diagnostic interface that gates every step on clinician judgment and turns each case into calibrated learning, and a quality-assurance agent whose employees train an “inexperienced colleague” toward autonomy, freeing capacity for judgment work. A survey and two independent LLM fidelity evaluations validate all ten patterns while exposing an actionability gap and a perception-enactment gap invisible to self-report.
As interface-level companion of the HI Manifesto series, the paper shows the complementarity definition is buildable and governable, and, by operationalizing judgment and problem decomposition, aims to move high-complexity problems such as protein-structure prediction from compute-only solvability into a hybrid regime where human structuring substitutes for raw scale, grounding a human-touch marketplace and durable human contribution at the high end of the complexity spectrum.
Keywords: hybrid intelligence; human-AI interaction; design patterns; participatory design; deskilling; future of work; LLM-based evaluation
Suggested citation: Sherson, J., van Bekkum, M., Bendler, J., Berthelsen, M., Boeg, J., Bozorgtabar, B., Bressendorff, M. J. R., Broholm, C., Buch, L., Catalan, C. R., Dahl, A. R., Dehler, K., Dellermann, D., Domanti, U., Elpelt, J., Esteves, J., von Felten, N., Fernandez, S., Filipova, M., … Zhang, Z. T. (2026). The Hybrid Intelligence Design Playbook: Operationalizing a decade of human-centered AI into interfaces where the human learns more than the system. Working paper, Center for Hybrid Intelligence, Aarhus University. Retrieved from https://hi-infinity.ai/research/the-hybrid-intelligence-design-playbook/