Designing Human-AI Experiences

Six load-bearing principles. One full day. Your product on the table.

For over twenty years, I have been designing at the intersection of human behavior and emerging technology. The hardest design problem of this decade is not how to use AI tools — it is how to ship AI products that people understand, trust, and choose to come back to. The field has been quietly converging on the answer for years; most teams just have not had the time to read it all.

That is what this workshop is for.

Across one full day in Berlin, you learn the 6 HAX Principles — Empathy First, Automation vs. Augmentation, Transparency and Confidence, Real Control and Editability, Graceful Failure, and Mental Model Shaping — and apply each one to a real product you ship or use weekly. The principles are not invented from thin air; they are researched, synthesized, and curated from the canonical work at Microsoft, Google, Apple, GitHub, the Shape of AI catalogue, AIverse.design, and the academic literature on human-AI interaction — pressure-tested across international fieldwork at Optimizer in Pharma, Healthcare, Finance, Logistics, and the public sector.

We use the 7 Sins of AI Product Design as the diagnostic lens and the 6 HAX Principles as the antidote, for shipping AI behavior — not just AI interfaces. Hands-on. Group work. A redesign and a ready-to-use prototype prompt by the end of the day.

You leave with three concrete artifacts: an audit of a real product, a HAX-principle redesign of its worst finding, and a prototype prompt ready for Figma, v0 or Lovable that takes the unhappy paths seriously — graceful failure, uncertainty signaling, real human agency.

For senior UX designers, product managers, design leads, and engineers shipping AI features. No machine-learning background required.

What you'll learn

The 6 HAX Principles, applied. Empathy First. Automation vs. Augmentation. Transparency and Confidence. Real Control and Editability. Graceful Failure. Mental Model Shaping. The six load-bearing beams of every AI product worth coming back to — applied, in the room, to a product you actually ship.

HAX Design Patterns. The practical patterns that make the principles concrete — explainability cues, calibrated trust signals, feedback loops that compound, error handling that does not insult the user, and editability as a first-class design problem.

Workshop schedule: 09:00 to 16:00, followed by networking & wrap-up until 17:00

09:00 – 10:25 · Module 1: The State of AI Product Experience

We open with a short talk on why the blank prompt box is a poor starting point for AI products, and what the alternatives look like in production. Then every participant tries the 13 prompt-free interaction patterns on hax.academy, hands on, and the tables compare notes: which pattern they would steal, and how many words they did not have to type. We close the module with the six HAX principles in ten minutes and the map that connects every pattern to a principle.

10:25 – 10:45 · Coffee Break

Morning refreshments and networking.

10:45 – 12:30 · Module 2: The Six HAX Principles, Explored

Two exploration rounds. Each participant works through one of six interactive principle demos and explains it to the table. Then the tables move beyond the chat box: eight voice patterns and eight camera patterns, with audio examples and a camera simulator on the phone. A table swap closes the morning: people change tables and compare what surprised them and which principle their own product fails most.

12:30 – 13:30 · Lunch Break

Networking lunch

13:30 – 15:00 · Module 3: The Audit, the Before

Each table picks one real AI product, ideally one somebody at the table ships. Using the 7 Deadly Sins of AI Product Design and the HAX Audit tool, the tables audit the product's behaviour, not its screens. Every finding gets a verdict and an evidence line. Three tables present their worst finding to the room, followed by an open discussion: who found the same thing in a different product?

15:00 – 15:15 · Afternoon Break

Quick refreshments before the final sprint.

15:15 – 16:00 · Module 4: The Redesign, the After

A short responsible-AI checkpoint first: fairness, transparency, accountability, privacy, and the two EU AI Act duties every team will meet. Then the tables redesign their worst finding with the patterns from the morning: a sketch of the After, at least two named patterns, one sentence ("Instead of typing __, the user now __"), and a prototype prompt generated by the audit tool, ready for Figma, v0 or Lovable.

16:00 – 17:00 · Networking & Wrap-up

Every table posts its Before and After on the wall; everyone walks and votes for the redesign they would ship tomorrow. The three most-voted tables present from the wall. We close with the take-away sheet, the 30-day plan, the scoreboard and a prize for the winning table, and four minutes for your feedback.