Start with a clearly bounded task
Hospital logistics robots transport items such as linen, meals, consumables or secured containers. Some move a built-in compartment; others tow or carry compatible carts. Autonomous mobile robots generally use onboard sensing and a map to navigate, while other guided vehicles depend more heavily on prescribed routes or infrastructure. The useful distinction is how each system behaves in your building, not its marketing label.
Begin with a repetitive route that has predictable demand and a clear owner at each end. Record who packs the load, requests the trip, receives the delivery and handles exceptions. A robot cannot solve an unclear receiving process. If staff must search for a delivery or repeatedly leave care tasks to unlock it, travel automation may simply relocate the workload.
Survey the route in real conditions
Corridor width is only the beginning. Assess turning spaces, thresholds, slopes, floor surfaces, crowded intersections, temporary equipment and doors that behave differently after hours. Observe patient transport, wheelchairs and walking aids. The robot must coexist safely with people whose movement may be unpredictable and whose hearing or vision may be limited.
Lift and access-control integration can become the critical path. Establish who authorises interfaces, how the system responds to a failed request and whether fire or emergency modes take priority. Charging locations must not obstruct escape routes. Wireless coverage should be surveyed along the whole journey, including service corridors and lift lobbies, rather than inferred from a signal reading at a desk.
Match security to the cargo
A clean linen delivery and a medicine delivery have different requirements. Locked compartments, authenticated access and an auditable handoff may be necessary for sensitive loads. Specimen transport can introduce packaging, temperature, spill response and chain-of-custody requirements. Confirm suitability with the relevant clinical, pharmacy, laboratory and infection-prevention teams instead of treating every container as equivalent.
Define which loads must never share a trip and how the robot and carts will be cleaned. Check material compatibility with approved cleaning products. For time-sensitive or temperature-controlled items, validate the complete journey, including waiting and failed-delivery scenarios. A robot’s rated payload does not establish that its compartment is suitable for a particular clinical cargo.
Calculate completed deliveries, not top speed
Useful capacity depends on loading, travel, lift queues, unloading, charging and recovery from blocked routes. A high advertised speed may have little value in a busy corridor where the system must slow or stop. Request route-level performance information and define when a task counts as complete: arrival at a waypoint is not necessarily a successful handoff.
Measure peak demand separately from the daily average. Include charging reserves and maintenance when sizing a fleet. Look for dispatch priorities, safe queueing, clear status messages and an easy way for staff to report a problem. The fallback process should identify who retrieves a stopped robot and how urgent deliveries continue without waiting for remote technical support.
Include the invisible integration costs
Request separate costs for vehicles, carts, docking stations, fleet software, connectivity, lift interfaces, access-control work and support. Clarify subscription terms, security updates and the consequences of losing a cloud connection. Facilities, IT and cybersecurity teams should review responsibilities before a contract is signed, including data access and remote maintenance privileges.
Compare the proposed service with the actual current workflow. Count staff minutes spent loading and recovering deliveries, not just walking time removed. Avoid presenting every saved minute as a cash saving: fragmented time may improve responsiveness without reducing a staffing requirement. Benefits can include more predictable supply movement and fewer interruptions, but these need local measurement rather than assumed percentages.
Run a pilot that includes bad days
Set acceptance criteria before the pilot: completed handoffs, intervention frequency, delivery timeliness, damage, near misses and staff burden. Include busy periods, temporary obstructions and planned network-loss tests conducted under an approved safety plan. Ask staff receiving deliveries what changed in practice. A smooth demonstration on an empty corridor answers only a narrow question.
Assign an operational owner, a technical owner and an escalation route. Establish maintenance checks, map-change control and rules for opening new routes. Expansion should follow evidence that the initial service is safe and useful. For environmental hygiene tasks rather than deliveries, read our disinfection robotics guide; the operating risks and validation requirements are substantially different.