When AI Gets Physical

Six Principles for Physical AI

At Punchcut, we’re designing physical AI across robotics, environments, and autonomous mobility. This work is teaching us that physical AI requires new principles that govern how intelligent systems behave, communicate, and relate to people. Here are six principles that that have been guiding our approach.

1. Signal intent. People need to know what a system will do before it does it. We solved a version of this problem for cars with turn signals, and drivers still read body language and interpret behaviors—drifting toward the lane line, slowing down at an intersection. Machines need their equivalent. Pausing before taking an action or moving. Decelerating on approaching. After the fact, intent is just a surprise.
2. Respect human space. Proximity, speed, orientation, and boundaries carry meaning and always have. How close is too close depends on the room, the task, and what person is standing there. We would experience a machine that treats a hospital corridor like a warehouse aisle as rude. This is a design decision, not a path-planning default.
3. Calibrate autonomy. There is a range autonomy across observing, suggesting, asking, acting, and standing down and the right spot on this continuum moves with the stakes, the system’s confidence, and who is present. Every product I have used that has gotten this wrong makes people uncomfortable—acting when it should have asked or asking about things nobody cared about.
4. Preserve human agency. People need an obvious way to interrupt, redirect, or take back control, and they need to know the appropriate behavior before they actually need it. Knowing you can stop the machine is what makes the fact that you probably never will bearable.
5. Communicate through all behaviors at once. Movement, sound, light, voice, and gesture. Each of these says something. When these behaviors do not align, people believe the movement. A cheerful voice from a machine that is crowding you does not read as friendly. It reads as oblivious or even intrusive. We must compose all these channels of communication together—just as we would for a visual system.
6. Earn trust slowly. First impressions matter, but consistency is what holds—that is, the same behaviors in the same situations, honest uncertainty instead of false confidence, or a graceful recovery when something goes wrong. One good recovery earns more trust than a hundred smooth interactions. One violated boundary can end it.

Designing for Shared Space

Underneath these six principles sits a question: how social should these machines be? People read intent into anything that responds to them. We do it with pets and vending machines, and we will certainly do it with something that turns to look at us. So, whether we mean to or not, we are designing a robot’s perceived qualities: confident or hesitant, attentive or indifferent, polite or pushy. The obvious move might be to make these systems seem human. However, that would be wrong. Faces and simulated warmth promise an understanding that the system does not have, and people quickly feel this gap. What a robot needs is enough social intelligence to behave appropriately around people. To yield in a doorway. Wait for someone to finish talking. Know the difference between a hallway and a bedroom. This is not about personality. It is about manners, which are a design problem.

Another thing that changes is who counts as a user. Physical AI encounters coworkers, children, guests, bystanders who have agreed to nothing, pets, and other machines. Most of the people it affects will never open the app or read the settings. Our unit of design is no longer one person and one product. It is a person, a system, and an environment full of others who have never opted in.

The Next Dimension of UX Design

We are good at asking what a system can do. The harder question is how an intelligent machine should behave around people. That pressing question requires UX design. Not because we need to make machines feel more human, but because we need to make their behavior understandable, appropriate, and accountable.

It also means expanding what we think of as a user interface. A machine’s movements, proximity, timing, uncertainty, and decisions all communicate something. Whether the machine waits, approaches, interrupts, or stands down is part of the user experience.

For many years, we’ve designed the space between people and screens. Physical AI asks UX designers to design the relationship between intelligent systems and people and the environment around them. While this is a bigger responsibility, it also presents an exciting opportunity to expand the impact of UX design.

Physical AI won’t succeed on its capabilities alone. It will succeed when intelligent machines can operate in the world in ways that people can understand, predict, and live with. Figuring out what this looks like is a job that UX designers cannot avoid.