The AI Manifesto

What does sustainable human-AI collaboration look like?

AI can produce content in greater volume and faster than any of us can read, understand or evaluate it. Good collaboration means work that is built around the pace, needs and limits set by human biology. Humans bring direction, judgment and responsibility to the work – AI brings speed and scale.

We value

• Developing understanding over the volume of output.
• Insightful questions over an abundance of answers.
• Focus and attention over parallel busyness and rush.
• Psychological safety and a sustainable pace over a fear-driven race.

While the items on the right have their place, we value the items on the left more.

Principles

1. Think carefully about what you ask for. Clarify what you want to find out and how you will evaluate the answer before you hand the task to AI.
2. Don’t accept output you don’t understand. If you can’t explain what the output does and what it is based on, it isn’t ready to use yet.
3. Respect the limits of your attention. Start only as many parallel tasks as you can genuinely manage, and rest when your attention tires.
4. Pause before deciding. A moment’s pause before a decisive step is a guarantee of quality, not a delay.
5. Make review work visible. Evaluating and correcting AI output is real work, and it often weighs on the team’s most experienced experts.
6. Trust with care, and verify. AI is brilliant at some tasks and convincingly wrong at others. Learn to recognise which one you are dealing with.
7. Challenge agreement. If AI always agrees with you, tell it to look for errors or blind spots in your thinking. Ask for opposing perspectives.
8. Maintain your own skills. Do things without AI from time to time. Skills that are not practised inevitably fade.
9. Don’t compete – grow. If everyone in a workplace simply tries to outrun the others with the machine’s help, everyone ends up exhausted.
10. Set your own limits. AI never tires and never tells you when the workday is over. You have to draw clear limits around your work, however tempting it is to keep going.
11. Speed is not direction. The subjective feeling that work is getting faster is no guarantee of a real increase in results.
12. Nurture human relationships. Difficult decisions, tolerating uncertainty and giving genuine feedback belong to the interaction between people.

These principles say what to do. The Research page explains why.

About

The AI Manifesto is written and maintained by Ari-Pekka Skarp – psychologist, psychotherapist and engineer. He has worked in technology organisations since 1999 and with Agile principles for more than 20 years. He has also helped people build better working lives as an occupational health psychologist, coach and psychotherapist.

The values and principles of the manifesto grow out of what people in organisations are experiencing right now as AI changes their work. Alongside that experience, they draw on a steadily growing body of research. The principles are updated as our shared understanding of the relationship between humans and AI grows.

Further reading

On my blog, Fractal Sauna:

• The 10-bit bottleneck
• AI does not make knowledge work easier – it raises the bar
• The AI revolution is burning employees at both ends
• The magic mirror that does not dare to say “Snow White”
• The tokenization of work and the wisdom of burnout