Inbound
Email intake
A dedicated mailbox with S3-backed ingestion captures emailed reports and attachments, parsing narratives and documents automatically.
dedicated mailboxattachments
Platform · AI Case IntakeAI-assisted case
AI-assisted case
intake with people in
control
Capture adverse-event reports from every channel and let AI accelerate extraction, MedDRA coding and data entry — while qualified personnel review, edit and approve every case before it proceeds.
GxP-alignedE2B(R2) / E2B(R3)21 CFR Part 11
1Multi-channel intake
Adverse-event reports captured from every channel into one queue.
email · web · APIInput
6 inbound channels
Email, portal, messaging, file & API feeds.
2AI-powered extraction
Structures patient, product, event & reporter; drafts the narrative.
domain-segmentedAI · Coding
MedDRA auto-coding
Events mapped to MedDRA PT / LLT terms.
AI · Scoring
Confidence scoring
Every field scored against a 0.85 threshold.
3Human-in-the-loop review
Qualified PV personnel review, edit and approve every AI field.
reviewer sign-offFlagged
Low-confidence routing
Fields below threshold forced to mandatory review.
4Case created
An approved, structured case enters the workflow, fully audit-trailed.
workflow-readyOutput
Downstream workflow
Assessment, follow-up and regulatory submission.
AI-acceleratedHuman reviewCase complete
How AI-assisted extraction works
From an unstructured report to a structured, scored draft case
AI reads the inbound report, identifies the regulatory-relevant entities, codes them, and assembles a draft case — attaching a confidence score to every field. High-confidence items are pre-filled; anything below threshold is routed straight to a qualified reviewer.
Inbound report · case AE-2026-04817email
from j.okafor@clinic.examplereceived 09:42 UTClang ENattachment discharge-summary.pdf
To whom it may concern, I am reporting on behalf of a treating physician. A 68-year-old female patient was started on Velmora 50 mg once daily on 14 Mar 2026.
Approximately ten days later the patient developed jaundice and severe right-upper-quadrant pain, with onset 24 Mar 2026. Liver function tests were markedly elevated. The patient was hospitalised for monitoring; the suspected medicine was withdrawn.
The reporter considers the event possibly related to the medication. The patient is recovering at the time of this report.
Suspect drugEvent / seriousnessDatesPatient / reporter
Extracted case datadraft · not submitted
Each field is extracted from the source and scored. Confidence below the configured threshold (0.85) is flagged for mandatory human review before the case proceeds.
PatientFemale, 68 yrs0.97
Suspect drugVelmora 50 mg once daily · oral0.96
Adverse eventHepatocellular injury MedDRA PT 100198370.91
Onset date24 Mar 2026 ~10 days after start0.94
SeriousnessSerious — hospitalisation0.95
CausalityPossibly related reporter assessment — verify0.71
1 field below threshold— routed to qualified review. Regulated judgments stay with people.
Draft case assembled — queued for qualified reviewAI accelerated the data entry; it did not submit. A reviewer confirms, edits and approves before the case enters the workflow.
6 fields extracted~40s from receipt
0
Intake channels
email · web · messaging · file · API
0.00
Confidence threshold
configurable per tenant & case type
0%
human-approved
a person signs off every case
0%
audit-trailed
every AI step & edit versioned
Multi-channel case intake
Cases arrive through the channels your operation already uses
No retraining your reporters. PVgenix ingests adverse-event reports from every relevant source, applies configurable intake forms and validation rules per tenant, and feeds them all into the same AI-assisted extraction pipeline.
Consumer
Messaging & social
WhatsApp, SMS and other consumer messaging platforms, plus social-media listening intake for adverse-event screening at scale.
WhatsApp / SMSsocial listening
Portal
Web reporter portal
Structured intake forms with HCP, patient and consumer-facing variants — guiding each reporter to the right fields for their role.
HCPpatientconsumer
Documents
File-based intake
PDFs and structured documents are parsed for case data — including scanned discharge summaries and CIOMS forms uploaded in bulk.
PDFCIOMSbulk upload
System
API & E2B feeds
System-to-system intake via API and E2B(R2)/E2B(R3) gateways, so partner systems and safety databases exchange cases directly.
REST APIE2B(R2)E2B(R3)
Per tenant
Configurable forms & rules
Intake form fields and validation rules are configurable per tenant, so each client's required data and quality checks are enforced at capture.
field configvalidation
Review · case AE-2026-04817MRM. Reyes · PV
AI-generated fields
Suspect drug · Velmora 50 mgconf 0.96 · accepted as extractedApproved
Adverse event · Hepatocellular injuryconf 0.91 · MedDRA PT confirmedApproved
Causality · Possibly relatedconf 0.71 · reviewer edited & confirmedEdited
Regulated judgments — stay with people
Seriousness determinationrequires qualified PV sign-offReviewer only
Expectedness & reporting clocknot an automated decisionReviewer only
Edit fieldsApprove & create case
Human-in-the-loop by design
AI assists. Qualified people decide.
AI accelerates intake and data entry — it does not make final regulatory or safety determinations. A qualified user reviews, edits and approves AI-generated content before any case proceeds, and every AI-assisted step is captured in the audit trail.
Review, edit, approve
Every AI-generated field is presented for confirmation. Reviewers accept, correct or override before a case is created — nothing proceeds unattended.
Regulated judgments stay with people
Causality, expectedness and seriousness sign-off remain with the client's qualified PV personnel — never automated.
Complete audit trail
Every extraction, confidence score, edit and approval is versioned and attributable — who changed what, when, and why — for inspection readiness.
This describes the platform's overall capability set, not a fixed shipped feature list; channels, thresholds and straight-through scope are defined per client agreement.
See AI-assisted case intake in action
Request a demo to see multi-channel intake, AI extraction, and human-in-the-loop review on your case types.