All case studiesProduction system
Lunalta.ai: menopause guidance answered from clinician-validated content.
Women going through menopause need reliable answers between appointments, and clinicians repeat the same guidance every day. Lunalta.ai, built and run by SIEL, answers only from clinician-validated content and hands anything outside that scope to a person.
- Client
- Lunalta.ai
- Sector
- Healthcare, women's health product
- Engagement
- Production system
- Measured result
- Liveproduct in use
Liveproduct in use
Growth goal
Give women going through menopause reliable answers between appointments, and give clinicians their repeated questions back.
Operating constraint
Guidance came in short appointments and long waits, and clinicians answered the same questions all day.
Measured result
A live product with a doctor-validated knowledge base and every consequential answer traceable to its source.
Summary
- Challenge
- Between appointments, women turned to search engines and forums and had to judge which advice to trust.
- What SIEL built
- Retrieval over a knowledge base curated and approved by the medical team, with answers composed only from retrieved passages. A portal for clinical oversight.
- Controls
- Scope limits define what may be answered. Out-of-scope or consequential questions go to a person, and every consequential answer cites its source.
- Impact
- Clinical standards and content stay with the medical team; SIEL builds and runs the system that applies them.
Lunalta had to be safe before it was clever. SIEL built the guardrails first: every answer traceable to content our doctors validated, and a clear handover to a person when it matters. They still run it with us.
The full account
01Reliable answers without a lower clinical standard
Questions about symptoms, sleep, mood and treatment do not wait for the next appointment. Clinicians, meanwhile, repeat the same standard guidance every day. Lunalta.ai was built to answer in those gaps and give clinicians time back. The condition was that nothing it says falls below the standard a clinician would stand behind, which ruled out answers from a model's general knowledge.
02Answers composed only from validated content
The medical team curates and approves every item in the knowledge base, and each item keeps its source. For each question, the system retrieves the relevant validated passages and composes an answer from them alone. If retrieval finds nothing that covers the question, the assistant does not fill the gap from general knowledge. It says so and points the user to the right next step.
03Scope limits and escalation
Scope limits define which topics the assistant may answer. A question outside that scope, or one that needs a clinician's judgement about the individual, is not answered by the model. It is escalated to a person, and the user is told what happens next. The medical team owns those limits and decides when they change.
04Clinical oversight through the portal
Clinicians use a dedicated portal to oversee patient interactions, see which content each answer drew on, and validate, update or withdraw content. A change applies to every later answer, so a correction is made once. Escalations appear in the same place and show which questions the assistant hands back to people. The product is live, built and run by SIEL, and every consequential answer traces to its source.
Discuss Your Operation With Our Engineers.
Describe one workflow and the systems it relies on. A senior engineer responds within two business days with an initial assessment: what we would build, what we would not, and why.