LGPDBrazilian Data Law — sensitive data
Health data is a sensitive category. Requires specific legal basis, prior DPIA, granular ACL and minimum retention. We deploy with default redaction and access logs ready for ANPD.
Applied AI for payers, healthcare.
Clinical triage, EHR, authorization and regulatory management — without compromising privacy.
Overview
What we deliver in this vertical.
Payers · Hospitals · Healthtechs
Healthcare
Clinical triage, EHR, authorization and regulatory management — without compromising privacy.
−45%technical denials
+3xtriage speed
99.2%coding accuracy
Applications
- 01Assisted triage and prevention
- 02NLP-driven EHR with ICD coding
- 03Authorization and bill audit
- 04Humanized support at scale
- 05LGPD and regulator compliance
Positioning
Healthcare AI doesn't replace the doctor — it gives time back.
The real gain in healthcare AI isn't blind automation — it's giving clinical professionals time back for cases where human judgment matters. Coding, denials, triage and chart structuring eat hours that should go to patients. Well-designed AI runs the bureaucracy in reverse.
By the numbers
What changes when AI enters the clinical workflow.
−0%technical denials in payers with automatic cross-check
+0×outpatient triage speed, AI-assisted
0%ICD-10/11 coding accuracy with human review
−0%structured chart time per encounter
Under the hood
A stack designed for clinical data.
Health data has specific regulation (LGPD + ANS + hospital guidelines) and requires isolation by default. The composition below is what shows up in mid/large hospitals and payers.
Models
- Self-hosted (Llama/Mistral)Default in environments with identifiable clinical data.
- Claude with BAAWhen latency requires a large model.
- Specialized models (BioBERT, ClinicalBERT)In medical NER flows.
Hospital integration
- HL7 FHIR APIStandard for TASY, MV, Soul MV, Wareline.
- DICOM gatewayImaging exams with per-study isolation.
- Custom webhookFor legacy systems without API.
Privacy
- Automatic PII/PHI redactionBefore any model call.
- Per-physician/department ACLNo data crosses boundaries without permission.
- Signed access logsFor internal LGPD/ANS audit.
Compliance
What regulates healthcare AI — and what changes in architecture.
Healthcare in Brazil has 3 overlapping regulators (ANS, ANVISA, ANPD). We treat each as a design constraint, not an end-of-line approval. POC without legal clearance doesn't reach production.
ANS RN 305ANS Normative Resolution
Standardizes payer operations. When AI enters authorization, coding or denials, it requires decision traceability and documented human review capability.
ANVISA RDC 657Software as a Medical Device
AI making clinical decisions (diagnosis, dosage, conduct) enters SaMD classification. We track development cycle and technical documentation from the first prototype.
TISSHealthcare Information Exchange
Mandatory standard for data flow between payer, provider and patient. Models consuming TISS respect original vocabulary and cardinality.
CFM 2.314Telemedicine and the medical act
Defines limits of the medical act in digital environments. AI assists but does not replace — every clinical decision needs a registered medical professional in the flow.
In production
Results in healthcare.
Technical denials · payer
Automatic cross-checking between EHR, contract and invoice. Technical denials fell 45% without increased provider friction.
Outpatient triage · hospital
AI-assisted triage accelerated prioritization 3× and reduced cases lost to delay. Doctor reviews every critical case.
ICD coding · regional payer
Automatic EHR coding to ICD-10/11 with human review. 99.2% accuracy measured against bi-weekly gold-set.
04 · LA AI Method
From hypothesis to operation.
Five steps that prevent eternal pilots. Each stage has deliverables, metrics and decision gates.
M1
Diagnostic
Opportunity map, estimated value, data readiness.
2–3 wksM2
Strategy
Prioritized roadmap, reference architecture, governance.
2 wksM3
Proof of Value
POC with production criteria and business metric defined.
4–6 wksM4
Deployment
Model, integrations, security, observability and UX.
6–12 wksM5
Continuous Operations
Evolution, fine-tuning, eval suite, cost per use, new features.
RecurringFrequently asked
About AI in healthcare.
How do you treat sensitive health data?
Brazilian data law + ANS guidelines. In hospital environments with clinical data, default to self-hosted models, identifier redaction and per-user ACL.
Does it work with hospital systems like TASY or MV?
Yes. We integrate via HL7/FHIR API or directly on the system DB, depending on what the customer prefers and the vendor allows.
Does it replace the medical auditor?
No. It elevates the auditor to decisions only on edge cases — automates the obvious ones and frees human time for cases where clinical judgment is essential.
ANVISA compliance for medical device?
When the use case enters medical device classification, we align dev cycle and documentation with ANVISA requirements. Not all healthcare AI needs this.