Almost anything, in theory — but ease drops fast with dexterity, force, judgement, and live sensing.
In theory, almost anything in a lab can be automated with enough time and money. In practice, automation works best for stable, repeatable tasks with rigid objects, clear visual targets, and simple success checks. It gets hard when a task needs dexterity, force control, judgement, or real-time sensing.
An automation-friendly task is one where the robot can clearly identify the input, perform the action, and verify the output. The more sensing and interpretation required, the more complex the automation. Simple automation moves plates, pipettes liquids, loads instruments, or follows fixed timing. Hard automation involves deformable materials, animals, fragile samples, unclear visual states, or steps where an experienced scientist is making judgement calls.
Loading a plate is straightforward — rigid object, defined location, checkable success. Sectioning tissue with a razor is much harder: the motion is simple but the tissue deforms, tears, shifts, and needs visual judgement. Mouse studies are harder still, demanding fast vision, precise force control, and protection of the animal.
A simple test: could a person do the step with one hand, in thick gloves, with limited vision? If yes, it's probably easy to automate.
If a step needs careful touch, live judgement, subtle visual interpretation, or fast reactions, it may still be automatable — but treat it as a custom engineering problem, not a standard equipment purchase.
