Flows B — B2B journeys
Teleconsult (patient + doctor + AI) · SOAP + CIE · Employer bulk enrolment · Insurer shared-savings reconciliation.
① Teleconsult · patient · doctor · AI co-pilot
Both sides see context. AI is ambient, never interrupts the conversation.
01
Pre-consult · doctor
primed
Aisha Al Mansouri · 14:00
Triage: chest tightness · 12h
Twin flags: HR pattern ⇢
CIE: no duplicate MRIs in 30d ✓
Draft agenda (3 items)
Doctor gets a 30-sec pre-brief. Triage + twin + claims history, all source-linked. Draft agenda editable.
→
02
In the call
ambient transcript
patient · video
◉ Recording · consent given
Live transcript · AR ↔ EN
CIE whisper: "ECG indicated per MOHAP pathway"
AI co-pilot whispers to doctor (private). Patient sees transcript, not prompts. Translation live between Arabic ↔ English.
→
03
Orders · one surface
cross-check
Orders
◉ ECG · in-clinic today
◉ Troponin · stat
○ Stress echo (hold)
CIE: pre-auth auto-eligible · Daman
Orders click-to-file into eClaimLink. Pre-auth predictions shown up-front. CIE flags guideline concordance, duplicates, leakage.
→
04
Patient gets · a letter
not a form
From Dr. Fatima
Aisha — thank you for calling. Based on what we discussed and your twin pattern, I'd like an ECG and troponin today. Here's the clinic nearest you…
◉ Book ECG · Jumeirah · 16:10
Plain-language summary replaces the clinical letter. Clinical version available in one tap. Available in AR, EN, UR, TL.
② Doctor SOAP · with Clinical Integrity Engine
SOAP auto-drafted from transcript; doctor edits; CIE checks before sign-off. The engine is a peer, not a cop.
01
Draft SOAP
AI-drafted
S · chest tightness 12h · radiates L arm
O · HR 88 · BP 128/82 · ECG: NSR, no ST changes
A · r/o ACS · likely MSK
P · troponin, NSAID trial, f/u 48h
Drafted from transcript. Every claim is citation-linked back to the conversation — doctor can hover to see source. Editable.
→
02
CIE nudges
pre-submit
3 nudges — review before signing
◯ NSAID · pt on metformin · no flag, but noting
◉ Per MOHAP guideline, troponin x2 (0h + 3h)
◯ ICD code I20.9 suggested · unstable angina?
Nudges, not blocks. Doctor can ignore with one click + reason. The reason is logged but not judged.
→
03
Guideline trace
why we think this
Why troponin x2?
MOHAP Acute Chest Pain Pathway 2024, §3.2: "serial troponin at 0h and 3h improves sensitivity for NSTEMI"
Evidence: ESC 2023 · quoted
Expandable on any nudge. Doctor can see the guideline + evidence. Never "the AI says so".
→
04
Sign → NABIDH
compliance
Submit SOAP
◉ File to NABIDH
◉ Attach to eClaim (auto-coded ICD-10)
◉ Send patient letter (EN)
Audit log · signed 14:28
One action = NABIDH file + eClaim + patient letter. ICD-10 auto-coded with confidence; doctor confirms. Timestamped, immutable.
③ Employer bulk enrolment
HR adds 500 employees, none are enrolled without their consent.
01
HRIS sync
one click
Connect HRIS
◉ Workday · OAuth
○ BambooHR · SCIM
○ Bayzat · CSV
518 employees detected
SCIM or OAuth preferred. CSV fallback. HRIS mapping screen allows HR to exclude (contractors, etc.).
→
02
Comms plan
HR-authored
How employees hear
◉ Email · personalized
◉ Slack · channel post
◉ WhatsApp · opt-in
Template editable · AR + EN
Invitation is personal, never mandatory. HR can A/B message. Subject line always mentions "voluntary".
→
03
Employee consent
from their device
Aisha — your employer offers Averra
Your company pays. Your data stays yours. Your HR sees aggregate stats only.
◉ Try it · free
○ Decline
Boundary shown · what employer sees
Hard boundary, visualized. HR never sees individual data. Declines are untracked past the aggregate.
→
04
Rollout board
HR dashboard
Rollout · Day 14
Invited · 518
Accepted · 368 (71%)
Engaged 30d · 298
n ≥ 20 everywhere · safe
HR sees rollout funnel, not who is who. Cohort views require k ≥ 20. Laggards reachable via broadcast only.
④ Insurer shared-savings reconciliation
Quarterly calculation: baseline vs actual vs attribution. Methodology is part of the contract.
01
Period close
Q2 2026
Q2 close · 30 Jun
Eligible cohort: 44,821 lives
Exclusions: new joiners <90d, catastrophic cases
Frozen snapshot · hash ◆
Cohort + exclusions snapshotted with hash. Both parties can audit the exact rows used.
→
02
Baseline vs actual
PSM-DiD
Methodology
Propensity-matched controls (n=44,821)
Difference-in-differences · 95% CI
waterfall · mini
PMPM Δ: −AED 38 · [−46 … −31]
Propensity-score matched difference-in-differences. CIs always shown. Spec locked by contract, not a moving target.
→
03
Attribution
what moved
Driver decomposition
Prevented admissions · 42%
Lower ED use · 28%
Generic substitution · 18%
Care-gap closure · 12%
Each driver has a confidence interval and a method link. Insurer can drill to the sub-cohort (k ≥ 20).
→
04
Invoice · two-party sign
both approve
Q2 invoice
Savings (attributed): AED 14.2M
Share to Averra (40%): AED 5.68M
Insurer signs ◉ · Averra signs ◉
Dispute window: 30d
Two-party sign + 30-day dispute window. Disputes trigger a data-room call with both actuaries. Nothing auto-bills.