Peer Hospitals Already Run Robot Fleets: Where Yours Stands

Are hospitals in New York and Florida really running robot routes?

Yes, though less uniformly than the headlines suggest. Rochester Regional Health reported in an article dated August 21, 2026 that its Moxi robots had passed 100,000 trips. In Florida, BayCare put two Relay robots, named Stork and Miles, live at Winter Haven Hospital and Winter Haven Women's Hospital on June 3, 2026.

Both figures come from the organizations involved, not from an independent measurement. That changes how you should use them. Treat the peer numbers as claims to test on your own floors, not as benchmarks to copy.

The three hospital count needs care too. Relay's BayCare case study describes the whole fleet as four robots across three hospitals, a cumulative description. The June announcement alone does not establish that count, so this article does not repeat it as a June expansion.

What Rochester reports, and how to read it

Moxi is built by Diligent Robotics. Rochester's program began in 2023, and pre-launch coverage described eight robots moving supplies between units at Rochester General and Unity, with an estimated more than 200 hours a week of work identified before launch. Staff named medications, laboratory specimens and patient belongings as the trips they wanted covered.

Since launch, the system reports more than 75,000 deliveries at Rochester General and more than 25,000 at Unity. It estimates more than 50,000 staff hours returned and about 55,000 miles of walking saved, roughly 120 million steps. Those are system estimates, not independent measurements.

Keep the two kinds of number apart. A trip count shows the robots ran. A staff hour estimate shows what someone believes those trips released. Your leadership will ask about the second, so build your own measurement before you repeat either.

What the Florida deployments show so far

BayCare's Relay work has the sharpest published figure. Relay reports a 99.8% delivery success rate at a St. Anthony's pilot running 50 deliveries per day. That number is vendor reported and describes a pilot, not the full Winter Haven fleet.

BayCare is also testing patient transport. A pilot with Rovex at Morton Plant Hospital, reported in April 2026, had not moved any patients yet, and its focus was workflow evaluation. That is an honest stage marker: the hospital is still learning how the work flows.

At HCA Florida JFK Hospital in Atlantis, Relay placed two robots, Dash and Daisy, to carry specimens, medications and equipment, mainly on time sensitive trips. Coverage dates the announcement to late October 2026. Staff counts, cost and measured results were not in the material we found, so JFK is a deployment to watch rather than a benchmark. Relay describes LiDAR, sensors and computer vision navigation, with authorized staff loading a locked compartment and choosing a destination.

Which of your runs are still manual?

Peer deployments cluster on a few run types: specimens, medications, supplies and patient belongings. Before comparing anything, write down every recurring run in your building using the same fields. Most hospitals find trips that were never tracked, often ones handled by whoever happened to be nearby.

Specimen runs need their own line. Custody and timing matter more than distance, so a run that looks short on a floor plan can still pull a licensed person away from bedside work for longer than the map suggests.

• Origin and destination, down to floor and wing
• Frequency by shift, and whether the run is time sensitive
• Elevator or door dependence, with typical wait time
• Who walks it today, and whether that person carries anything else on the same trip
• What happens when the run is late or missed

What staff time do those runs consume?

Count it, do not estimate it from memory. Pull timestamps from a representative sample of recent weeks, and separate walking from elevator waits, handoffs, and the checking that confirms a specimen or bag actually arrived. Walking time alone understates the cost of a manual run.

Peers headline staff hours, but the method usually sits in a footnote. Rochester's 50,000 hours is a system estimate spread over the life of the program. Do not convert a peer figure into your savings, because floor layout, elevator banks and staffing will shift the result.

Your target is a number a finance lead can trace back to specific runs. That is a shorter sentence than a benchmark, and a more useful one.

What a short, measured pilot can test first

A pilot should answer one question: does this route hold up on this floor, across these shifts? Pick one run type and one wing. Run it long enough to cover your slowest shift and your busiest one, and write the stop criteria before the first trip, not after.

Keep one manual route running alongside the robot. A parallel comparison is far more reliable than a before and after snapshot, because demand on a hospital floor shifts from week to week.

• Completed deliveries against requests, counting failures and manual rescues
• Time from request to pickup, and from pickup to drop off, median and worst case
• Elevator and door hold events, logged by location
• Staff touches per run, including loading and unloading
• Exceptions that needed a person, and how long recovery took

Where Service Robot Co. fits in

Service Robot Co. is an OEM neutral integrator for US businesses. We choose robots across manufacturers for the route in front of us, then deploy, integrate, train and service each unit under RoboCare, our service plan, backed by regional service engineers in every state we service. The robots named in this article come from Diligent Robotics and Relay Robotics, and whether either brand fits your floors is a question for a route test, not a brochure.

We offer robots by lease only. A three year lease carries one fixed lease payment, quoted after a site assessment that measures the route, elevator access and charging needs first. That keeps the number tied to your building rather than to a catalog page.