IPG Health

AI Playbooks

Regulated Content Systems for Global Pharmaceutical Markets

Problem
$51M portfolio across 13 brands (Novartis 6, Sanofi 7) with fragmented compliance workflows and 30% rejection rates
Solution
Unified content governance platform with automated compliance checks, achieving zero HIPAA violations over 4 years
AI Playbooks Platform
$51MPortfolio managed
13Pharma brands
0HIPAA violations
60%Cost reduction

$51M portfolio. 13 brands. Zero room for compliance failures.

IPG Health managed pharmaceutical marketing for Novartis (6 brands including Kesimpta) and Sanofi (7 brands). Each brand had different MLR requirements, regional variations, and approval workflows.

I embedded with both client teams and internal production staff. Mapped existing workflows, identified bottlenecks, and cataloged compliance failures. The pattern was clear: 30% of submissions were rejected not for content quality, but for process violations.

Teams were using spreadsheets to track approvals, email chains for reviews, and shared drives with inconsistent naming. No single source of truth. No audit trail. No way to demonstrate compliance during inspections.

30%rejection rate
28day avg cycle
13brands to manage
Diagnose Problem

Compliance Status by Client — Before AI Playbooks

30% of content was non-compliant or pending review, causing delays and rework across both portfolios.

$0M$31M$61M$92M$122MNovartis - compliant: $72MNovartis - nonCompliant: $28MNovartis - pending: $15MNovartis$115MSanofi - compliant: $68MSanofi - nonCompliant: $32MSanofi - pending: $22MSanofi$122MCombined - compliant: $70MCombined - nonCompliant: $30MCombined - pending: $18MCombined$118M
compliant
nonCompliant
pending

One platform. 13 brands. Complete compliance isolation.

AI Playbooks is a multi-tenant content governance platform. Each brand operates in isolation—separate workflows, separate approval chains, separate audit logs—but shares common infrastructure and compliance rules.

Architecture priorities: tenant isolation for compliance, automated MLR checks before submission, complete audit trails for regulatory inspections, and workflow automation to reduce manual handoffs.

Technology Stack

Backend: Python/FastAPI, Node.js
Database: PostgreSQL, Redis
ML/NLP: spaCy, BERT (claims extraction)
Search: Elasticsearch
Infrastructure: Azure AKS, Terraform
Monitoring: Datadog, PagerDuty
CI/CD: Azure DevOps, ArgoCD
Storage: Azure Blob, S3
Architect Solution

AI Playbooks Architecture Layers

7-layer architecture providing multi-tenant compliance governance with HIPAA-compliant data handling.

AI Playbooks Architecture

Multi-tenant content governance platform for regulated pharmaceutical markets

1
Presentation Layer
2
API Gateway
3
Orchestration Layer
4
Compliance Engine
5
Content Services
6
Data Layer
7
Infrastructure

Compliance Engine Architecture

Automated validation pipeline that reduced rejection rates from 30% to 8%.

1
Claims Extraction
NLP identifies product claims
F1: 0.94
2
Library Matching
Compare against approved claims
> 0.85 similarity
3
Fair Balance Check
Risk/benefit ratio validation
85% auto-pass
4
Regional Rules
Country-specific requirements
12 regions
5
Template Validation
Correct format for content type
95% auto-pass
Validation Results by Check Type
Check TypeAuto-PassAuto-FailManual Review
Claims Validation72%8%20%
Fair Balance85%3%12%
Template Format95%4%1%

99.9% uptime. Zero HIPAA violations. Four years running.

AI Playbooks runs on Azure AKS with multi-region failover. Every component is monitored. Every transaction is logged. Every deployment is automated and tested.

SLOs that matter: sub-500ms response times for all user interactions, 99.9% uptime with automatic failover, complete audit logs retained for 7 years (regulatory requirement), and zero data breaches.

The platform processes 500+ assets per month across all brands. Each asset goes through automated compliance checks, workflow routing, and approval tracking—with full visibility for stakeholders at every step.

99.9%uptime SLA
500+assets/month
380msp95 latency
Engineer Solution

Platform OKRs — Current Quarter

Key objectives and results tracking cycle time, scale, and reliability targets.

9
Complete
0
On Track
0
At Risk
0
Behind
Reduce MLR Cycle Time
Production LeadDirector
89%
Average review time < 14 days
12 / 14 days
86%
First-pass approval rate > 75%
78 / 75 %
100%
Revision rounds < 2
1.6 / 2 avg
80%
Scale Multi-Brand Operations
Platform TeamEngineering
NaN%
Platform Reliability
SRE TeamInfrastructure
72%

60+ staff. 13 brands. One way of working.

Rolled out AI Playbooks across the entire portfolio: content producers, MLR reviewers, brand managers, and client stakeholders. Each role got tailored training and workflows optimized for their responsibilities.

Established playbooks for every content type: promotional materials, medical education, sales training, patient materials. Each playbook includes templates, compliance checklists, and approval workflows.

Before

  • ✕ Spreadsheet tracking
  • ✕ Email-based approvals
  • ✕ No audit trail
  • ✕ 28-day cycles

After

  • ✓ Automated workflows
  • ✓ In-platform approvals
  • ✓ Complete audit trail
  • ✓ 12-day cycles
Enable Solution

Platform Adoption Trends — 12 Month View

Key metrics showing platform adoption and operational improvements over time.

Assets Processed
18.5%
540
Min: 180Max: 540
Avg Cycle Time
42.8%
12d
Min: 12dMax: 28d
First-Pass Rate
51.9%
79%
Min: 52%Max: 79%
Team Satisfaction
43.8%
4.6/5
Min: 3.2/5Max: 4.6/5

The Result

Four years of operations. Zero HIPAA violations. Zero compliance failures during audits. The platform became the standard operating model for regulated content across both Novartis and Sanofi portfolios.

MLR cycles dropped from 28 days to 12 days. First-pass approval rates increased from 52% to 79%. Infrastructure costs dropped 60% through cloud migration and automation.

Most importantly: the team scaled from 5 to 60+ while maintaining compliance standards. The playbooks and workflows I established became institutional knowledge—surviving team changes and client transitions.

28 days12 daysMLR Cycle Time
52%79%First-Pass Rate
5 staff60+ staffTeam Scale
Impact

Annual Value Creation — $4.8M Savings (56% Reduction)

Breakdown of operational cost reduction through automation, cloud migration, and process optimization.

$0M$50M$100M$150MBaseline Costs: +$85M$85MMLR Cycle Savings: $-12M$-12MFirst-Pass Improvement: $-8.5M$-8.5MInfrastructure (60%): $-21M$-21MCompliance Automation: $-6.5M$-6.5MFinal Costs: +$37M$37MBaseline CostsMLR Cycle SavingsFirst-Pass ImprovementInfrastructure (60%)Compliance AutomationFinal Costs
Total
Increase
Decrease

Lessons Learned

What Worked Well
  • Starting with compliance engine before workflow automation—front-loading quality reduced downstream rework
  • Tenant isolation from day one prevented cross-contamination during audits
  • Investing in audit trail infrastructure early made regulatory inspections stress-free
What We'd Do Differently
  • Would have built claims library integration earlier—manual claim lookup was a bottleneck for 6 months
  • Should have implemented regional rule engine as configuration, not code—each new region required deployment

"The playbooks outlasted three reorgs and two client transitions. That's the real measure of institutionalized knowledge."