Summary

• Enact Intelligence emerged from stealth with a $30 million pre-seed round backed by three venture firms.
• Its technology collects force information through actuators rather than relying only on camera-based perception.
• The company plans to expand its engineering team and work with industrial partners, with validated deployment results still outstanding.

Zurich robotics startup Enact Intelligence has emerged from stealth with $30 million in pre-seed financing to develop machines that can sense forces through their actuators, pursuing industrial manipulation tasks that remain difficult to automate reliably with cameras alone.

The round attracted venture investors redalpine, Notion Capital and firstminute capital, with individual backing reported from Mistral AI co-founder Arthur Mensch, AI researcher Yann LeCun and Path Robotics founder Andrew Lonsberry. Founded in March 2026, Enact is led by Manuel Yves Galliker, previously involved in the NEO humanoid programme at 1X and robotics research company Physical Intelligence.

Enact is developing an integrated approach spanning AI models, data pipelines, actuators and sensors. Rather than adding a tactile accessory to an otherwise established robot, the company wants physical feedback to shape how its machines learn, move and respond to contact with objects.

Its proposition is technically plausible but remains at an early commercial stage. The company has not released independently validated production performance figures, a detailed product price or evidence that its approach is more economical than established industrial automation. The funding announcement establishes the resources behind development, not a completed industrial deployment.

Enact’s emergence reflects growing investment in physical AI, where the reliability of a machine depends on interactions with objects and environments rather than the quality of text or image outputs. A robot that can recognise a component still needs to handle it without slipping, crushing it or damaging adjacent equipment.

Force information changes the control problem

Industrial vision systems can identify the apparent position and orientation of an object, but an image alone does not reveal how much resistance a gripper is encountering or whether a connector is beginning to bind as it enters a socket. These details can determine whether an assembly operation succeeds.

Enact’s proposed architecture collects force-related information through actuators, the components that generate robotic movement. By relating motor behaviour to physical resistance, a system can obtain information about its interaction with the environment without depending exclusively on external tactile sensors.

The company describes a proprioceptive approach. In biological systems, proprioception provides information about body position, movement and exertion. A robotic equivalent could help the controller estimate how the machine’s joints and mechanisms are behaving even when camera visibility is limited.

Consider fitting a delicate connector. Visual guidance may bring the components close together, but small misalignments can produce resistance before either object visibly moves. A force-aware controller could detect that difference, reduce pressure and adjust its path rather than continuing a motion that risks damage.

Achieving that response consistently requires fast and reliable measurements, suitable control software and training data that connects actions with physical results. Noisy actuator readings or an incorrect interpretation of resistance could lead a robot to stop unnecessarily or apply force in the wrong direction.

The promise of integrating models with hardware is that the system can be optimised around those measurements. The difficulty is that changes to motors, gears, sensors or mechanical geometry can alter the data available to the model, complicating development and production repeatability.

Industrial use demands repeatable results

Factories already employ robots for welding, handling, packaging and assembly, frequently within carefully engineered environments. Fixtures and programmed movements reduce variability, allowing machines to perform repetitive tasks reliably at high speed.

More adaptable robots could extend automation to work that is currently performed by people because components vary, processes change frequently or manual intervention is needed when something does not fit as expected. Food handling, maintenance and selected manufacturing tasks present potential opportunities, although their requirements differ substantially.

A demonstration of dexterity is not the same as production readiness. Industrial buyers must establish successful task completion rates, cycle times, equipment life, maintenance requirements and the cost of integrating a new machine into existing operations.

A force-aware robot might handle unfamiliar objects more successfully than a fixed programme but still deliver a poor return if it works slowly, requires costly retraining or frequently calls for a specialist to recover from errors. Production decisions depend on throughput and reliability over long periods, not a small number of successful demonstrations.

Safety imposes additional constraints when a robot interacts with workers or expensive equipment. Machine limits, emergency stops and independent protective systems must remain effective even if an adaptive AI controller encounters an unfamiliar situation or makes a poor prediction.

Enact’s reported twelve-person team draws experience from companies and research organisations including 1X, DeepMind, Amazon Robotics, Helsing and Physical Intelligence. The financing is expected to support substantial recruitment by the end of 2027 and accelerate hardware and research work with prospective industrial partners.

Recruiting across mechanical engineering, electronics, control systems and machine learning will be necessary because the product requires coordinated development across all those disciplines. It is also expensive, particularly before the company has demonstrated repeatable manufacturing and service processes.

Zurich provides access to established robotics research and engineering expertise, while the founders bring experience from American AI development. Neither advantage guarantees a commercial product, but both may help the company work across the physical and software components of its design.

The next credible milestones will be documented industrial trials comparing Enact’s machines against existing processes on speed, reliability and total operating cost. The pre-seed round demonstrates strong investor appetite for tactile robotics; whether actuator-based sensing can improve real-world production remains to be established.