What was announced
M42 announced Ossi, which it calls its next-generation multi-agent AI Clinical Platform, at the Abu Dhabi Future Health Summit. The company says it is built on Microsoft Azure's cloud and AI infrastructure. The stated aim is to cut administrative burden and give clinicians more time with patients. It also aims to bring data and evidence-based guidelines to the point of care.
Gulf Business reported the launch alongside M42's plans for a new Bahrain facility. It described Ossi as already activated across seven M42 healthcare facilities in the UAE and Bahrain.
The seven sites, per the Abu Dhabi Media Office release:
- Amana UAE
- Amana Bahrain
- Healthpoint
- HealthPlus
- Mubadala Health Dubai
- Sheikh Sultan bin Zayed Hospital
- Imperial College London Diabetes Centre in Abu Dhabi
The release does not say what "activated" means in practice. It gives no figures for clinicians or patients using the system. It also doesn't say whether every department at those sites is covered.
The four agents
Ossi launches with four agents, which M42 describes as the first in an expanding suite. The release describes them as follows:
| Agent | What it does, per M42 | Human review stated? |
|---|---|---|
| Ambient Scribe | Transcribes clinical conversations and prepares draft documentation | Yes, clinician review |
| Health Record | Retrieves and summarises relevant EMR information | Not specified |
| Clinical Q&A | Helps clinicians find clinical information and prepare patient-friendly educational content | Not specified |
| Coding | Suggests medical codes from the documentation | Yes, review by the appropriate teams |
Every output is framed as a draft, a summary or a suggestion. Nothing in the announcement claims autonomous diagnosis, treatment decisions or final coding. The release names human review only for the scribe and coding agents. It describes no approval or audit process for the other two.
Architecture: what is stated
For administrators, the integration claims matter most:
- Model-agnostic: M42 says the platform works with different large language models rather than being tied to one. It names no models.
- Multilingual: The release gives no list of languages.
- EMR integration: It connects to existing electronic medical record systems through HL7 and FHIR standards, or through custom APIs. It names no EMR vendors and no HL7 or FHIR versions.
- Speech and language stack: Kareem Shaheen, M42's Group Chief Digital Officer, said the platform combines LLMs with speech recognition and natural language processing. He also claimed real-time EMR synchronisation. That is an executive statement, not a technical specification.
M42's own product page for Ossi says the platform is built on M42's proprietary architecture and leverages Microsoft's AI infrastructure. The press materials do not say Microsoft built the agents. Microsoft's role is as cloud and AI infrastructure provider and collaborator.
Why it matters for Microsoft-focused readers
This sits within an existing relationship. In 2024, Microsoft said G42's data platform and other essential technology infrastructure would migrate to Microsoft Azure. M42 is a G42 subsidiary. Microsoft's healthcare AI model catalog has also included M42 work. M42 was among the partners that contributed foundational models to the Azure AI Foundry catalog.
Ossi also fits a wider pattern of clinical documentation tools running on hyperscale cloud. A hospital group builds an orchestration layer on a cloud platform, keeps the models swappable, and connects to the EMR through standard interfaces.
Elena Bonfiglioli, Microsoft's Global Business Leader and VP for Healthcare, Pharma and Life Sciences, called Ossi an important milestone in the collaboration. She said Azure provides a trusted foundation for it. That is a vendor endorsement of the platform's hosting. It is not a statement that Microsoft has audited or certified the system.
Security and privacy
The announcement says M42 worked with Microsoft to ensure Ossi meets stringent health data security and privacy standards. It also says the platform supports compliance with regional data governance rules, including the UAE's. Gulf Business similarly reports that M42 says Ossi supports regional data governance requirements and includes safeguards for patient information security and privacy.
The release leaves out several things a hospital IT or compliance team would want to see:
- named certifications or specific regulations
- data residency arrangements
- encryption and access controls
- retention policies
- whether patient data is used for model training
- results of any independent audit
Treat the compliance language as a marketing-level assurance until M42 publishes those details.
What is missing
The release offers no accuracy or error-rate data for transcription or summaries. It gives no safety study, no measured time savings and no evidence that patient interaction has increased. The claim that this is "a big win for both patients and clinicians" is M42's characterisation. The same goes for CEO Dimitris Moulavasilis's remarks about a shift toward predictive, personalised and proactive care. Those are forward-looking aims, not features of the four agents now deployed.
Questions worth asking as deployment continues:
- How often do clinicians have to correct the scribe's draft notes?
- How are hallucinated or omitted details caught before a note is signed?
- How are suggested codes audited?
- How does multilingual and mixed-language speech perform in real clinics?
- Who is accountable when an EMR sync fails or a summary misses a critical history item?
Bottom line
Ossi is a real deployment announcement rather than a pilot teaser. It is live at seven facilities across two countries. It has four defined agents, and it runs on Azure with HL7/FHIR integration. Its effectiveness, safety record and compliance posture are still unproven in public. For IT teams weighing similar projects, the architecture is the useful part: a model-agnostic layer, standards-based EMR connections, and human review built into the drafting and coding steps.