The keynote
The Copilot Never Lands the Plane. 45 to 60 minutes, three interactions, six research slides with the source on every one. For associations, conferences and company events.
The copilot never lands the plane.
AI does more of the flying every month. I help entrepreneurs and knowledge workers stay pilot-in-command: keeping their judgment, their values and their joy in the work.
Research-based. Myth-free. Practical.
You are probably worse at judging your AI-assisted work than you think. So am I.
That is not an opinion. It is what the best research of the last two years shows.
The better you think you are with AI, the worse you judge your own results.
In two experiments with about 500 participants, people solved logic problems better with ChatGPT and overestimated how well they had done. The participants with the highest AI literacy judged themselves least accurately. Source: Fernandes et al., Computers in Human Behavior, 2026.
AI is built to agree with you.
Across 11 leading AI models, AI affirmed what users did 49% more often than humans did. In three preregistered experiments with 2,405 people, a single conversation made people more convinced they were right. And they preferred the model that flattered them. Source: Cheng et al., Science, 2026.
Human plus AI is often worse than the better of the two.
A meta-analysis of 106 experiments found that combinations of humans and AI often performed worse than the best of human or AI alone on decision tasks. They did better on creating content. Source: Vaccaro, Almaatouq & Malone, Nature Human Behaviour, 2024.
The bottleneck in working with AI is not prompting skill. It is knowing what you know, and noticing when the tool is steering you.
The Brainfulness Cockpit: four layers, one rule.
The Brainfulness Cockpit is a four-layer framework for working with AI: the flight plan, the pilot, the copilot and the decision. It has one rule: the copilot never lands the plane.
Flight plan. Why you are flying at all: your values, your purpose, what makes the work meaningful. The layer most often skipped when tasks go to AI.
Pilot. You, and your brain, with its instruments, habits and blind spots.
Copilot. The AI. Three questions tell you how far to rely on it: is it able, what is its motive, is it stable?
Decision. Who owns the result, what an error costs, and who may say "I'm not comfortable with this."
And one habit that ties it together: HOLD, a cross-check of 5 seconds, 30 seconds or 5 minutes.
Three ways to work with me
The Cockpit workshop
Half a day, up to 25 people. Participants run the cross-check on their own tools and their own decisions, and leave with the checklists they filled in themselves.
The Airline Day
One day, one organization. Keynote and workshop for staff, a Tower session for leadership and HR, and a measurement before and 30 days after.
A lecturer, not a motivational speaker.
My job was always the same: take something complicated and make it simple enough to use on a Tuesday morning. I now do that with the research on humans and AI.
I am not a neuroscientist and not an AI researcher. I read the research, check where the numbers come from, and turn what survives into something you can do. Expect examples, metaphors and jokes, mostly at my own expense.
Every number on this site has a source you can check.
Most statistics about AI and burnout that pass through your LinkedIn feed come from consultancy surveys, and some cannot be traced to any source at all. On stage and on this site I use peer-reviewed research as proof. Consultancy surveys appear only as stories, with the sender named. If you find a claim here without a source, tell me and I will fix it or remove it.
Planning an event about AI and the people who work with it?
Tell me the date, the audience and what you want them to do differently the next morning.