LeanTaaS

The optimisation engine is the product. Forecasting demand from historical patterns, recent shifts and real time signals, then computing a better allocation of operating room blocks, infusion chairs or beds, is not something a scheduler can do by hand at this scale, and there is no record system or network asset underneath doing the real work.

Two things hold it below A. The company describes itself as selling software AND services, and its origin is a management consulting methodology applied to healthcare, so a meaningful part of the value is the deployment and change management around the algorithm. And the mechanism is described as patented optimisation combined with lean principles and simulation, which is operations research as much as machine learning. That is a strength rather than a criticism, and it does mean the artificial intelligence framing is looser here than for a model driven vendor.

The documented behaviour is decision support. The system forecasts and recommends, and schedulers, surgeons and nurse managers act on the recommendation, with the stated aim of helping staff act quickly and decisively rather than acting for them.

The product named Autopilot is where the question sits and it is unanswered. Launched in 2024 as the first generative artificial intelligence product for hospital operations, its name implies action taken without a person, and nothing published describes what it does unattended, what it may change on its own, or what approval sits between a generated recommendation and a moved surgical block. A block release or a chair reallocation is a real world action affecting a scheduled patient, so the distinction matters.

The technique is named honestly rather than dressed up: patented optimisation algorithms, lean principles, predictive and prescriptive analytics, simulation methodology. Describing simulation and operations research alongside machine learning is more accurate than most vendors in this index manage.

Patents are an unusual and genuine transparency channel, because a granted patent is a public document describing a method in enforceable detail, which is more than a marketing page ever contains. What is absent is performance: no forecast accuracy, no measure of how far the optimised schedule sits from an achievable optimum, and no description of how the models are retrained as a hospital's demand pattern shifts.

An unusual and genuine disclosure channel carries this record. The technique is named honestly rather than dressed up, covering patented optimisation algorithms, lean principles, predictive and prescriptive analytics and simulation methodology, and describing operations research alongside machine learning is more accurate than most vendors in this index manage. The patents matter more than the naming does.

A granted patent is a public document describing a method in enforceable detail, examined by an office that required the description to be enabling, so it is a form of technical disclosure no marketing page can match and one a technically capable buyer can actually read. This index should credit it wherever it appears. What it does not do is describe the operating chain.

No model or model family is named for the learned components, no hosting arrangement is published, and no sub processor list was located, and no position on whether customer scheduling data trains models serving other hospitals was found.

That last question has the same shape recorded elsewhere in this index for products whose value comes from a cross customer comparison base: the more hospitals contribute demand patterns, the better the forecasting gets, and whether one hospital's data improves a competitor's schedule is the material term. Ask for it, and for a sub processor list.

The independent validation is the strongest available in this category. KLAS named the company Best in KLAS for Capacity Optimization Management in 2026 with 95.6 out of 100, the highest score in its category and a repeat of the previous year, and those rankings are drawn from tens of thousands of provider interviews rather than from vendor submissions.

The deployment figures are specific and countable rather than rounded: nearly 200 health systems, more than 1,200 hospitals, over 5,600 operating rooms, 14,500 infusion chairs and 23,000 inpatient beds. Named academic customers include Stanford Health Care, UCHealth, UCSF and NewYork Presbyterian.

Held at B rather than A on a real distinction: KLAS measures customer satisfaction, not operational outcome. No published study quantifies the utilisation improvement, waiting time reduction or added case volume with a baseline and a period, which for a capacity product is the measurement that would settle the question.

Graded on an honest basis. No published stewardship position, retention schedule or encryption detail was located in this pass.

The data profile is different from most records here and worth understanding rather than assuming. This product consumes scheduling, throughput and utilisation data rather than clinical narrative, so the sensitivity is lower per record than for a documentation or diagnostic system. It is not zero: a surgical schedule identifies who is having what procedure and when, an infusion schedule discloses that a named patient is receiving chemotherapy, and the volume is enormous across 1,200 hospitals.

Graded on an honest basis and flagged for re verification. No compliance statement or agreement posture was located in this pass.

A vendor operating in nearly 200 health systems has necessarily executed business associate agreements at scale and passed the vendor risk review of large academic medical centres many times over, so the terms exist and are simply not public. That is a documentation gap rather than a substantive one, and it should be resolved by asking rather than inferred either way.

A search covering security certification terms returned nothing attributable to the company, and no trust centre, SOC 2 report, HITRUST certification or ISO certification was located. A dedicated trust search remains owed and this grade is provisional until it is run.

The same reasoning as the compliance axis applies and should temper any conclusion drawn from the absence. Deployment across nearly 200 health systems including several major academic medical centres means this vendor has cleared a great many hospital security reviews. Whatever it holds is not surfaced in public material.

No device pathway applies and none is claimed. Scheduling and capacity software does not diagnose, treat or support a clinical decision about an individual patient's care, so it sits outside device regulation cleanly, and its absence from that framework is correct rather than a gap.

The accountability question that does arise is not regulatory but contractual. When an optimisation system reallocates capacity and a case is delayed, responsibility rests with the health system that acted on the recommendation, and nothing published describes how the vendor and the customer divide that in practice.

Nothing published on model monitoring, subgroup analysis or governance, and the failure mode here is distinctive enough that its absence is the most substantive gap on the record.

This is an allocation system. It decides which surgical blocks are released, which infusion appointments are offered when, and how beds are assigned, across scarce resources that are by definition insufficient for demand. Allocation of scarce medical capacity is one of the most studied sites of inequity in health services research, and an optimiser tuned on historical utilisation inherits whatever the historical pattern encoded, including which surgeons held blocks, which service lines were prioritised and which patients previously got access. Optimising against that history efficiently reproduces it. Nothing published addresses whether the objective function considers equity of access at all, or only throughput and revenue.

Two passes located no forecast accuracy, no measure of how far an optimised schedule sits from an achievable optimum, no description of how models are retrained as a hospital's demand pattern shifts, and no warranty, indemnity or remediation commitment. What this product decides makes the absence consequential rather than routine, and this index has recorded the pattern under a broader thread: the system allocates a finite good.

Operating theatre time and infusion chairs are scarce, and a schedule is a decision about who is treated when, so a forecasting error does not produce a wrong document, it produces a patient waiting longer or a slot going unused. Neither outcome generates a complaint that reaches the vendor, because a patient scheduled three weeks out has no way to know a better schedule existed.

The optimisation objective is the thing to establish in writing, because efficiency, throughput and revenue are not the same target and the difference determines who waits. Nothing published states what the model optimises for, whether clinical urgency is a constraint or an input, or what override a scheduler retains. Ask for the objective function in plain terms, for forecast error against actuals, and for what the vendor commits to when a forecast is materially wrong.

Integration is a precondition rather than a feature here, and the deployment footprint is the evidence. The company states its products work in conjunction with existing record systems, and operating across more than 1,200 hospitals requires reliable access to surgical scheduling, infusion appointment and inpatient census data, each of which lives in a different module and often a different system.

A partnership with a major imaging and diagnostics manufacturer to optimise operational performance suggests integration reaches beyond the record into departmental systems. Held at B because no named certification, interface standard or write back mechanism was located, and because whether the system writes schedules back or only recommends them is not described.

Cloud based, stated plainly and not elaborated. No hosting region, customer controlled option or retention schedule was located.

One dependency is worth a buyer's attention. Capacity optimisation is only useful if it is current, so the data flow is continuous rather than periodic, and an outage during a working day removes the schedule optimisation from an operating suite that has already been planned around it. No availability commitment or continuity position was located.

No price, rate card or pricing mechanism is published, and no indication whether the model is per operating room, per chair, per bed or per facility, which matters because the product line is explicitly organised around counts of exactly those assets.

What is published is unusual in the other direction: the company discloses its own commercial position, reporting annual contract value approaching 150 million dollars and 170 percent revenue growth over three years. That is corporate transparency rather than pricing transparency and it does give a buyer something real, because dividing a disclosed contract value across a disclosed number of hospitals produces a defensible order of magnitude that most vendors make impossible.

Deep in hospital operations and deliberately bounded there. The three assets addressed, operating rooms, infusion centres and inpatient beds, are the constrained resources that determine throughput in an acute hospital, and coverage across nearly 200 health systems makes this one of the most operationally embedded products in United States hospitals.

The buyer is an operations, perioperative or capacity leader rather than a clinician, and the product is specialty agnostic in the sense that it manages the room rather than what happens inside it, while being tied to surgical, oncology infusion and inpatient settings specifically. Nothing addresses ambulatory clinics, emergency departments or anywhere outside the United States.