For oncology and clinical research
AI assists. People decide.
Specialized digital assistants for patients, physicians, researchers and coordinators. Not a universal chat interface, but a team of bounded agents around a real healthcare process.
Product system
Each assistant receives only the context required for its role, knows when to stop and hands a reviewable result to a person.
Patient and family
Clarifies the next step and helps people stay oriented between appointments.
Does not diagnose, prescribe treatment or assess urgency instead of a clinician.
Metriccompleted pathway · escalations · comprehension
Physician
Reduces routine work around a decision without making the decision.
Does not sign records, alter prescriptions or present unchecked text as clinical fact.
Metricdrafting time · corrections · acceptance
Researcher
Speeds up source review, protocol analysis and preparation of an auditable hypothesis.
Does not replace the statistician, ethics committee, principal investigator or peer review.
Metricsearch time · completeness · reviewer errors
Coordinator
Moves an administrative task forward while making every action visible.
Does not determine trial eligibility, interpret tests or hide automated actions.
Metriccycle time · incomplete tasks · touchpoints
Trust architecture
Every clinically material action passes through four visible layers.
Only authorized data and the user’s role
One bounded workflow and predictable actions
Sources, logs and stopping rules
Confirms the clinical or research decision
Market atlas · 18 Aug 2026
Solutions shown: 8
Turns clinician–patient conversations into reviewable draft notes linked to the source conversation.
A draft inside the workflow, traceability and mandatory professional review.
Scale and outcome claims come from the vendor and cannot be transferred to a local pilot without validation.
Supports draft clinical documentation and compatible EHR workflows.
Role-specific interfaces, workflow integration and confirmation before the record is finalized.
Microsoft limits the product to support for qualified professionals; it does not replace medical judgment.
Provides clinicians with literature-linked answers; Cedars-Sinai has reported EHR integration.
Evidence stays next to the answer and inside the professional workspace.
Evidence retrieval does not replace appraisal of study quality or clinical judgment.
Uses conversational symptom assessment to route users toward an appropriate level of care and local resources.
A clear dialogue, a localized pathway and an explicit handoff to care.
Each language, population and local pathway requires separate clinical validation.
Supports non-diagnostic tasks such as outreach, reminders, information collection and structured handoff.
Narrow roles, non-diagnostic tasks and a return path to the responsible organization.
Vendor claims about scale and safety require independent validation and do not justify high-risk expansion.
The official product page describes management of agents for medical and operational tasks.
Russian-language clinical context, specialized agents and process integration.
Functions, training-data scale and effects are vendor-reported; each clinical scenario needs independent local evaluation.
An NLM and NCI system for search, matching and ranking of clinical trials, evaluated in a pilot user study.
Search, criterion-level explanations and ranking for expert review.
Research results do not authorize autonomous screening or enrolment.
Operational assistants support outreach, task tracking, missing-information collection and coordination.
The agent advances an administrative task and maintains an exception queue.
Vendor performance figures require local comparative validation.
Publications
An English knowledge base on medical AI agents, clinical trials and oncology pathways, with dates and sources that can be checked.
Market and practice · 8 min read
How medical AI agents differ from chatbots, which workflows are being deployed and why clinical decisions remain with people.
Clinical trials · 9 min read
How AI can support oncology trial search and criterion matching without deciding eligibility or enrolment.
Patient pathways · 8 min read
A clear guide to AI-assisted navigation across Russia’s OMS insurance, VMP high-technology care and self-pay pathways—without treatment selection.
Implementation strategy
We reuse specialization, workflow integration, source traceability and human handoff. Outcomes are measured in the local pilot.
Track 00
Track 01
Track 02
Track 03
Proposed base pathways
Each pathway is designed as a separate agent workflow with traceable sources and a clear handoff.
Find current studies, show criterion-level rationale and hand the enquiry to a coordinator and physician.
Explain the official OMS pathway, questions for the physician and commonly required documents.
Explain referral grounds, VMP paperwork and the route to a specialist medical center.
Compare service scope, clinical assumptions, documents and cost transparency without selecting treatment.
Safety by default
Diagnosis, prescriptions, trial enrolment and other clinical decisions remain with an authorized professional.
Material outputs are linked to a document, version and reviewable passage.
The agent receives only what it needs. The public demonstration does not use real medical data.
Actions and escalations are recorded; automated steps should be visible and, where possible, reversible.
Start with one bounded task, an error set and stopping criteria, then seek expert evaluation.
The agent does not replace emergency services. In an emergency, contact local services; in Russia call 112 or 103.
Open design track
Propose one specific task for a patient, physician, researcher, coordinator or healthcare organization—without medical data or a promise of implementation.
Propose your agent ↗Review basis · 18 Aug 2026
