Where narrow autonomy is genuinely appearing today
Full autonomous surgery, a robot deciding and executing an operation without a surgeon actively directing each consequential motion, remains far off, but narrower forms of assistance are genuinely being explored and adopted in specific, bounded tasks. Image-guided positioning assistance is one of the clearest examples: a robot arm can use pre-operative or real-time imaging to help position an instrument or guide precisely to a planned trajectory, for instance in orthopedic procedures, while the surgeon still makes every decision about whether and how to proceed and typically still executes or closely supervises the actual cutting or placement. This works well because the task, hit this precise planned point along this planned path, is well-defined, geometrically constrained, and verifiable against the plan before anything happens.
Suturing assistance is another area seeing incremental autonomy, where a system might handle a repetitive, well-defined sub-step, like driving a needle through tissue along a planned path, once a surgeon has set up and approved the specific stitch, rather than the robot planning an entire suturing sequence unsupervised. Rehabilitation and exoskeleton robots are arguably the domain where autonomy is furthest along, because the task there is fundamentally different: assisting or resisting a patient's own movement during physical therapy according to a program, in a much lower-stakes, non-invasive, continuously human-supervised setting, rather than making irreversible decisions inside a body. These systems can adapt resistance and support in real time, essentially applying control-loop and actuator concepts covered elsewhere on this site to a rehabilitation context.
