Phasing a Hospital Robot Tender: Start With Three

How should a hospital tender allow many robots but start with few?

A hospital does not have to commit to a full fleet on day one. The cleanest structure is a tender that names a ceiling, such as 30 units, and funds an initial group, such as three. The rest sit as an option that opens only when defined conditions are met. That is the shape of the public tender Sheba Medical Center in Israel has been running, as reported by the vendor side of the story.

The first three units carry most of the decision. Pick the routes, the departments and the measures before the first robot leaves its dock. Those choices decide whether the data you collect can later justify expansion, or whether it simply describes three robots doing three jobs.

What does the Sheba tender actually show?

Gix Internet and its subsidiary Deliverz.ai report that Sheba's public tender permits up to 30 delivery robots, with 3 currently deployed. Deliverz.ai was selected under the tender after a successfully completed pilot. A tender ceiling is a potential, not a guarantee of future orders.

The headline result is that average chemotherapy medication delivery fell from about 40 minutes to about 14 minutes over 11 months. One company release describes this as roughly a 65 percent reduction, while the Deliverz.ai website gives 64 percent compared with manual delivery. Both figures are vendor data. Use them as a claim to test, not a benchmark to copy.

According to a Gix Internet announcement, commercial operations began in December 2025, and more than 9,000 autonomous missions had been completed as of October 2026. An earlier announcement cited more than 7,000. Mission counts climb quickly, so check the date before quoting any figure in a board memo.

How should you choose the first three units?

Start with routes where time is expensive and a late delivery has a visible clinical cost. Sheba's first robot work followed that logic. An earlier pilot by a different Israeli vendor ran drug-delivery robots on existing maintenance tunnels and elevators, starting in oncology. There, a nurse orders chemotherapy and the robot carries it. Sheba said it hoped to expand the system over time.

Score each candidate route before you commit. A route that looks short on a floor plan can run long once a robot waits for an elevator or queues behind a cart. Use these criteria to rank them:

• Repetition: trips per shift, so three robots see steady work rather than occasional runs
• Elevator dependence: number of lifts crossed and typical wait time at peak hours
• Clinical urgency: whether a late delivery delays treatment, a procedure or a lab result
• Corridor load: how many people, carts and beds share the path during busy periods
• Staff time now: hours nurses or runners spend walking the route by hand

What performance data should you collect?

Set the baseline before go live, not after. Time each manual delivery on the chosen route, from order to receipt, and then run the same route with a robot. Without that comparison, the first month of numbers proves very little.

Track more than average minutes. Averages hide the bad tail. Log failed or aborted missions, stops in corridors, elevator waits, and the share of trips completed without a person stepping in. Measure uptime across the whole fleet, not just the units that happen to be charged and ready.

Sheba's 40 to 14 minute result shows what a measured average can look like, but it reflects one medication route and one vendor's logging. Your floors will differ. Report the median and the slow end of the range, and record the staff hours recovered alongside the minutes.

When should a hospital expand beyond the first units?

Expansion should follow evidence, not the calendar. According to the Gix Internet release of October 12, 2026, Sheba's robots began carrying bone marrow samples from hematology to the laboratory, and donated blood from the donation center to the blood bank. The company says more medication and specimen routes are planned, and notes that this is not guaranteed.

That sequence matters. Each new workflow is a new route with its own map, baseline and clinical sign-off. A workable gate set for the optional units looks like this:

• Uptime holds at the level written into the tender across a full review period
• Delivery times beat the baseline on the original route, not only on its best days
• Failed missions fall, and each one has a logged cause and a documented fix
• Nursing and transport staff confirm that the time saved is real and usable
• The next route has its own map, baseline and clinical approval before it opens

How should the tender language protect the hospital?

Write the ceiling as an option. The contract should state that the hospital is not obligated to take units beyond the initial group. It should also fix the price of additional units at signing, so expansion becomes a decision rather than a renegotiation.

Ask about the software layer as closely as the hardware. Deliverz.ai describes its platform as one orchestration layer that coordinates autonomous robots and human transport teams, and says it is not tied to a single robot model. A hospital should ask who owns the route and performance data, how a robot from another manufacturer would be added, and what happens to scheduling if one vendor leaves the market.

Where does an integrator fit in a phased plan?

A phased tender is mostly an operations project. It needs people who can map the building, match the robot to each route, and keep the fleet running after the vendor demonstration ends. Service Robot Co. works as a full-service commercial robot integrator that is OEM neutral. We pick the right robots across manufacturers, then finance, deploy, integrate, train and service every unit through a nationwide US engineer network, so the hospital works with one company.

For a site starting with three robots, we offer a three-year lease with one fixed payment, quoted after a site assessment. The assessment covers the route mapping a phased plan depends on, including elevator and corridor conditions, before any unit is committed.