With OPTIMABIO, French university hospitals aim to make every prescription an AI-assisted decision.

The OPTIMABIO project aims to improve one of the most common decisions made in hospitals: ordering laboratory tests. Led by Assistance Publique-Hôpitaux de Marseille (AP-HM), Hospices Civils de Lyon, Limoges University Hospital and French startup Kiro, the initiative has received more than €17 million in funding through the France 2030 i-Démo programme.

At first glance, the project may appear less spectacular than an artificial intelligence system capable of detecting a rare disease or interpreting medical images. Yet its impact could prove significantly greater. Behind the optimisation of laboratory test orders lies a major shift in the development of the digital hospital: artificial intelligence is moving beyond specialised tools used for isolated tasks to become a core function embedded directly within hospital software.

The objective is no longer simply to produce a better diagnosis, but to improve, in real time, the thousands of clinical decisions healthcare professionals make every day.

Prescribing becomes AI’s next frontier

Unlike many other medical fields, laboratory medicine combines several characteristics that make it particularly well suited to artificial intelligence. Its data are highly structured, medical guidelines are extensively codified and best-practice recommendations are regularly updated. Above all, laboratory medicine informs nearly 70% of diagnostic decisions, giving it a central role across the care pathway.

Against this backdrop, the objective is to improve the relevance of every test order. The algorithm developed through OPTIMABIO will analyse available laboratory and clinical data in real time to identify the tests best suited to each patient’s situation. It could flag that an identical test has already been performed, recommend a more appropriate examination or highlight a discrepancy with established medical guidelines. AI therefore intervenes upstream in the decision-making process, without ever replacing the clinician.

This approach addresses a major economic problem. The OECD estimates that around 20% of healthcare expenditure is wasted on ineffective or inappropriate care. An unnecessary prescription does not merely represent the cost of an additional test. It consumes hospital resources, lengthens care pathways, places unnecessary pressure on laboratories and can trigger a cascade of further examinations. Reducing unnecessary orders therefore improves both the quality of care and the efficiency of the hospital system.

AI moves from standalone application to native hospital software function

The real shift introduced by OPTIMABIO lies primarily in how the technology is integrated.

Over the past decade, most digital health startups have marketed applications or platforms designed to sit on top of existing tools. This accumulation of software has often proved to be one of the main barriers to adoption, multiplying interfaces, disrupting user workflows and integrating poorly with hospital information systems.

OPTIMABIO takes a radically different approach. Its AI-generated recommendations are intended to be integrated directly into the software physicians already use every day. AI is no longer an additional product but a native capability within the hospital information system.

This development mirrors what has happened in enterprise software. After initially offering standalone assistants, major software vendors gradually embedded artificial intelligence directly into their ERP and CRM platforms and productivity suites.

University hospitals become co-producers of artificial intelligence

The project also reveals a less visible transformation in the role of university hospitals. For years, healthcare institutions were primarily regarded as testing grounds for startups. They supplied data, validated clinical protocols and trialled solutions developed by private companies.

OPTIMABIO points to a different model. The three university hospitals involved are not merely participating in an experiment. They are contributing their medical expertise, laboratory guidelines, prescribing protocols and the knowledge accumulated by their teams of medical biologists. In other words, they are directly involved in building the algorithm.

This shift is far from insignificant. As artificial intelligence advances, value no longer lies solely in the models themselves, but also in the data, operational rules and clinical expertise required to specialise them.

Hospitals are therefore becoming producers of digital assets, capable of turning medical knowledge into software intelligence.

France 2030 is funding infrastructure, not merely a startup

The €17 million awarded to the project under France 2030 is supporting an industrial capability designed for large-scale deployment across the French hospital system.

The objective is to build infrastructure capable of improving medical practice over the long term.

A strategy that sidesteps the main obstacles facing medical AI

Kiro’s positioning is also strategically significant.

Most medical AI startups have historically focused on applications directly related to diagnosis. These use cases involve high levels of responsibility, particularly demanding regulatory approval processes and strong resistance from practitioners whenever the technology appears to compete with their expertise.

By focusing on the appropriateness of laboratory test orders, Kiro is choosing a markedly different path.

The artificial intelligence does not decide between two diagnoses. It provides contextual recommendations based on medical guidelines and previously available results. The final decision remains entirely in the clinician’s hands.

This approach reduces medico-legal risks while making the technology easier for healthcare professionals to accept. AI becomes a copilot for clinical decision-making rather than a substitute for the physician.

The next battleground for hospital AI

OPTIMABIO is likely to open up a market extending well beyond laboratory medicine. If artificial intelligence proves capable of improving the relevance of laboratory test orders, the same logic could be applied elsewhere in the care pathway: medical imaging requests, treatment selection, patient triage, prevention of redundant examinations, scheduling of specialist consultations or optimisation of hospital discharge.

In other words, hospital AI could gradually move into clinical operations. Organisational decisions are far more numerous than complex diagnoses, and they also account for a significant share of hospital costs and the potential improvements in care quality.

Artificial intelligence could therefore create more economic value by optimising millions of everyday decisions than by automating a small number of highly specialised medical procedures.

A new generation of digital hospitals

Several challenges remain, beginning with data governance across institutions, interoperability with the many hospital software systems already in use, the explainability of recommendations and the project’s business model once public funding comes to an end. Each will be decisive.

These questions do not, however, diminish the paradigm shift embodied by OPTIMABIO. Hospital AI is entering a phase of normalisation, one in which it seeks to quietly improve the decisions that shape healthcare professionals’ everyday work.

This may be where its most transformative potential lies: no longer in its ability to occasionally rival a specialist on a complex diagnosis, but in its capacity to support, thousands of times a day, the routine decisions that determine the quality, speed and cost of patient care. OPTIMABIO illustrates a deeper trend: medical AI is ceasing to be an exceptional tool and becoming an invisible but essential component of the hospital’s daily operations.