ANiGENT

Detection is the whole product and machine learning is named as the method in the company's own voice, including in the launch announcement made jointly with Mayo Clinic. The formulation they use is a blend of applied human logic and advanced machine learning, which is more candid than the lane norm because it concedes that rules carry part of the load rather than presenting everything as a model.

It sits below the top grade for that reason and because two of the four modules, AUDITOR and Waste Reconciliation and Assay, are reconciliation and laboratory workflows that would function without a model at all.

This vendor makes the most aggressive claim in the lane and governs it the least. Its own product page states that the Peer to Peer module can predict, within minutes, which specific users will divert medications at certain locations and times. That is a claim to name employees before they have done anything, and it is qualitatively different from every competitor here, which surface behaviour that has already occurred for investigation.

Nothing published describes what a customer is meant to do with a prediction of future theft, what threshold generates one, who is qualified to review it, whether the named person is ever told, or what happens if they are monitored, reassigned or dismissed on the strength of it. A claim of that reach with no accompanying oversight model is the finding that carries this grade.

Mechanism disclosure is reasonable for a company this size. The method is named as a blend of applied human logic and advanced machine learning, five data sources are cited, Microsoft Power BI is identified as the visualisation layer, and the technical differentiator is stated concretely: modelling more than 300,000 combinations of how a medication can be handled, against the one to one dispense and administration relationship the company says competitors track.

That is a checkable architectural claim rather than a slogan. Holding it at this grade is the near 100 percent accuracy figure, which is offered with no definition of accuracy, no population and no method, and which pulls against the credibility of the disclosure around it.

No model provider, framework or third party analytics component is named for the machine learning itself. Microsoft Power BI is identified, but as the dashboard and visualisation layer rather than as any part of the detection model, and treating it as a supply chain disclosure would misread what it does. Following the acquisition nothing describes whether the models remain those built by ANiGENT or are being replaced by, merged with or fed into the parent's own analytics stack.

The evidence is specific, quantified and compromised in the same breath. A five year benchmark study at Mayo Clinic Arizona reports practice inefficiencies reduced to under one percent and more than $3 million recouped over the period, with more than 80 facilities on the platform as of 2022. Two things qualify it.

Mayo Clinic co founded the company alongside DYNALABS, so the flagship validation was conducted by a related party and should be weighed as one rather than read as independent confirmation. And what the study measures is revenue recapture and practice efficiency, which is not the same question as how reliably the platform detects diversion. Set against that, the separate claim of near 100 percent accuracy in identifying the true cause of a variance carries no definition, population or method and does not function as evidence.

Nothing published addresses whether customer transaction data trains models beyond the tenant, how behavioural profiles of named staff are retained, or what happens to them when a contract ends. The exposure is larger here than the vendor's size suggests, because the Peer to Peer module builds forward looking behavioural predictions about identified employees rather than only flagging completed transactions, and a predictive profile of a named person is a more sensitive artefact than an alert about a discrepancy.

The company site carries a HIPAA Statement as a standing footer document alongside its privacy policy, terms and legal notices, which is more published structure than several vendors in this lane maintain. The document itself was not retrieved in this pass, so its content is unassessed and the grade cannot rest on it. Recorded here as the specific artefact a future pass should retrieve first, since reading it is the single cheapest action that could move this grade. Beyond that link, no business associate agreement position, data ownership statement or description of how identified clinician and patient records are handled was located.

No certification, attestation, trust portal or security page was located in two passes. One point needs stating explicitly to prevent a wrong inference. The parent holds two HITRUST i1 certifications, but those are scoped by name to specified Omnicell products and the cloud platform hosting them, and MAAP Analytics is not among the products enumerated in that scope. A buyer cannot read Omnicell's certification as covering this platform, and the distinction is exactly the kind that gets lost when a small vendor is absorbed by a larger certified one.

Surveillance software rather than a regulated device, so no clearance applies and none is claimed. The launch material positions the platform as helping organisations prevent regulatory and legal risk, and the AUDITOR module produces reconciliation output that would serve an audit, but no statement of the company's own regulatory position was located.

Nothing published describes how its findings should be used in reporting to the Drug Enforcement Administration or a state board of pharmacy, which matters for a product whose output routinely becomes the basis of exactly those reports. The published positioning here is thinner than that of the larger competitors in this lane.

No governance statement, validation summary, false positive rate or appeal route was located, for a system whose headline capability is predicting which named individuals will steal drugs. Every bias concern this axis exists to catch applies here in its strongest form.

A model that forecasts future misconduct from historical behavioural patterns will concentrate its predictions on whoever the historical patterns already disfavour, which in a hospital means night shifts, high turnover units, agency staff and the least senior clinicians. Publishing nothing about who the model names, how often it is wrong, or whether the distribution was ever examined leaves the most consequential question about the product entirely unanswered.

No published position on responsibility, challenge or remedy. This is the lowest grade on the axis for the strongest reason of any record in this lane. Where competitors flag completed transactions for investigation, this product claims to predict which named individuals will divert in future, and an employee identified by such a prediction has done nothing that can be examined, disproved or explained.

Nothing published states who owns that determination, what weight it is intended to carry in an employment decision, whether the person is informed, or how a prediction is withdrawn. The more anticipatory the claim, the more recourse matters, and here the two run in opposite directions.

The platform is described as drawing on five key data sources including controlled substance transactions tied to patient records, and an explicit design goal is operating without intensive informatics resources from the customer, which is a meaningful commitment for a product sold to hospitals whose integration teams are the bottleneck.

The sources are not fully enumerated in public material, no electronic health record or dispensing cabinet vendor is named, and nothing describes the integration method, so the depth claimed cannot be verified. Five sources is also a narrower input surface than the larger competitors in this lane document.

The platform is described as on demand, which implies a hosted service, and nothing beyond that adjective was located. No hosting model, cloud provider, region, residency commitment or customer option was found in two passes. The acquisition by Omnicell in October 2025 makes this more rather than less pressing for existing customers, since nothing published states whether MAAP data has moved, will move, or remains where it was.

No price, pricing page or basis of charge was located across two passes. One commercial commitment is published and worth recording because it is unusual: the MAAP Analytics 3.0 release was delivered to existing facilities automatically and at no additional cost, with monitoring and support included through the transition, which is a stated position on upgrade charging that most vendors leave open. That does not reach the threshold this axis measures, which is whether a buyer can establish what the product costs before entering a sales process.

The published footprint is hospitals and health systems, with more than 80 facilities cited as of 2022 and a stated ambition to serve health systems in the United States and globally. A distribution partnership announced in 2022 with ConsortiEX, which supports hospital pharmacy compliance across a large site base, extends reach on paper.

What is missing is any breakdown by setting: nothing distinguishes acute inpatient from procedural, retail, ambulatory surgery or long term care deployment, and the operating room and anaesthesia areas that competitors call out specifically are not separately addressed in published material.