Normalized claim
Conversations evaluated: 926 conversations increase
Over a three-week period, we conducted a randomized controlled experiment with 926 cases
MicrosoftEvidence: Medium50/100
Alan deploys physician-supervised conversational medical AI
Use case typePatient engagementUpdated Nov 1, 2024
Alan, a health and insurance company operating in France, Belgium, Spain, and Canada, introduced an LLM-based conversational agent into its medical advice chat service in 2024. The agent helped patients ask medical questions, receive clearer guidance, and interact with physicians through a supervised, privacy-compliant workflow.
- Organization
- Alan
- Industry
- Healthcare
- Location
- France
- Published
- November 2024
Reported outcomes
926 conversations
conversations evaluatedOther quantified impact
298 conversationscompleted interactions4.6 /5patient satisfaction3.7 /4information clarity+95%physician positive ratings+81%opt-in rate among respondents
Strategic outcomes
Customer experience & trustImproved patient satisfaction and claritySpeed & agilityFaster back-and-forth patient conversationsRisk & complianceMaintained safety under physician oversightNew business modelAugmented existing medical advice chat service
Why do we believe this?Outcome claims, sources, and evidence checks
Normalized claim
Completed interactions: 298 conversations increase
handled 298 complete patient interactions
Normalized claim
Patient satisfaction: 4.6 /5 increase
overall satisfaction (4.58 vs 4.42 out of 5, p < 0.05)
Normalized claim
Information clarity: 3.7 /4 increase
higher clarity of information (3.73 vs 3.62 out of 4, p < 0.05)
Normalized claim
Physician positive ratings: 95% increase
95% of the conversations were assessed as “good” or “excellent”
Normalized claim
Opt-in rate among respondents: 81% increase
The high opt-in rate (81% among respondents) exceeded previous benchmarks for AI acceptance in healthcare.
Normalized claim
Patient response time: 61.1% decrease
patients also responded more quickly compared to control conversations (median: 1.1 vs 2.8 minutes, p < 0.001)
Normalized claim
Provider response time: 95.8% decrease
provider response times differed significantly (median: 0.2 vs 4.8 minutes, p < 0.001)
Why do we believe this deployment?Customer identity, provider attribution, maturity, and source checks
- Customer
- Alan
- Provider
- Microsoft
- Maturity
- Unknown
- Linked source
- arXiv
No explicit deployment-stage evidence found.
Customer identity supportedSource describes one deploymentMaturity evidence evaluated
Primary read
Use case focus
Showing 2 of 2
- 1Patient engagement
- 2Conversational assistants
- Alan faced rising demand for medical advice amid doctor shortages and wanted to improve access to medical expertise.
- The company needed an AI-assisted patient-facing workflow that preserved safety, privacy, and physician oversight.
- Alan integrated an LLM-based conversational agent into an existing medical advice chat service.
- The system used multiple sub-agents and models from OpenAI, Anthropic, and Mistral AI served by Microsoft Azure and Google Cloud Platform, with physician review, transparent AI labeling, consent collection, and staged rollout safeguards.
- Over a three-week study, 926 conversations were evaluated and 298 were completed with the agent.
- Patients reported higher clarity and satisfaction, and physicians rated 95% of conversations as good or excellent.
Architecture
Alan integrated a physician-supervised LLM conversational system into its existing medical advice chat service. The system used multiple sub-agents running in parallel, with models served by Microsoft Azure and Google Cloud Platform. Safeguards included explicit AI labeling, consent collection, restricted operating hours and scopes, physician review within 15 minutes, a staged rollout, and mandatory physician check-ins after conversations.
Sources & evidence1
Evidence: Medium50/100Evidence strength
- Customer explicitly identified
- Independent source available
- Quantified outcome available
- Technical implementation details available
Type: Research ReportPublished: Nov 1, 2024Publisher: arXivEvidence: SecondaryConfidence: Low
AI-generated summary. Verify important details with the linked sources before relying on this case.
Explore related AI use cases
Customer evidence
Provider evidence
AI patterns
Browse the catalog
Was this useful?
Community
Comments
No published comments yet.