Pfizer

CoCo

AI-Enhanced Intranet for Enterprise Knowledge & Onboarding

Problem
1,000 new hires across 20+ fragmented systems with 40% MLR rejection rates
Solution
Governed RAG platform reducing cycles by 65%, tripling throughput, saving $2.08M annually
CoCo Platform
65%Faster MLR cycles
Throughput increase
$2.08MAnnual savings
~1,000Staff onboarded

1,000 new hires. 12 weeks. 20+ fragmented systems.

Pfizer split its marketing business between IPG and Publicis right before peak vaccine season. Publicis inherited Paxlovid, Abrysvo, and Comirnaty—and needed to scale from zero to nearly 1,000 staff in weeks.

I embedded with Pfizer, Publicis, and IPG for two weeks before writing any code. Shadowed content producers. Sat with MLR reviewers. Mapped the actual workflow.

What I found: staff spent 25+ minutes per asset just searching for approved templates and prior claims. 40% of MLR submissions were rejected—not for content issues, but for using wrong templates or outdated claims.

25+min to find one asset
40%MLR rejection rate
20+disconnected systems
Diagnose Problem

MLR Content Pipeline — Before CoCo

Only 20% of content requests made it through to approval, with 40% rejected at first review due to template and claim errors.

Content Requests: 1.0K (100%)1.0KContent Requests100%Initial Draft: 720 (72.0%)720Initial Draft72.0%28.0%MLR Submission: 580 (80.6%)580MLR Submission80.6%19.4%First Review: 420 (72.4%)420First Review72.4%27.6%Revisions: 340 (81.0%)340Revisions81.0%19.0%Approved: 200 (58.8%)200Approved58.8%41.2%
Total Visitors:1.0K
Conversions:200
Overall Rate:20.00%

CoCo: an AI-enhanced intranet that answers questions and finds content.

CoCo—"Company Companion"—is embedded in Microsoft Teams. Instead of searching 20 systems, employees ask CoCo. It retrieves approved content, answers policy questions, and guides new hires through processes—with citations back to source documents.

Architecture choices: RAG layer over existing systems of record. Azure ML + AKS for the pipeline. Cognitive Search for hybrid retrieval. Graph-RAG for relationship mapping across brands and regions. MLR gateway for compliance checks before any content surfaces.

User asksRetrieveGenerateCite

Sources: Veeva Vault · SharePoint · Workfront · CLM · Regional Drives

Architect Solution

CoCo System Context (C4 Diagram)

The platform integrates with 6 core systems while maintaining a single conversation interface via Teams.

Loading diagram...

Security & Compliance Architecture

Defense-in-depth model with HIPAA, SOC 2, and FDA 21 CFR Part 11 compliance.

CoCo Security & Compliance ArchitectureHIPAA-compliant RAG platform with enterprise-grade data protection1Perimeter SecurityNetwork edge protectionAzure WAFDDoS ProtectionGeo-filtering2Network SecurityInternal network controlsVNet isolationNSG rulesPrivate endpoints3Identity & AccessAuthentication and authorizationAzure ADRBACMFA4Application SecurityCode and runtime protectionAPI validationInput sanitizationOWASP Top 105Data ProtectionEncryption and maskingAES-256 at restTLS 1.3 in transitPII redaction6Audit & MonitoringLogging and alertingAzure MonitorLog AnalyticsSIEM integrationCertificationsSOC 2 Type IIHIPAA BAAISO 27001FDA 21 CFR Part 11Security MetricsIncidents (12mo)0Audit Findings0Pen Test ScoreA+Compliance100%Defense-in-depth model: Each layer provides independent protection. Outer layers protect inner layers.

Disaster Recovery & Business Continuity

Multi-region architecture with automated failover and tested recovery procedures.

RTO
2.5 hours
Target: < 4 hours
Met
RPO
15 minutes
Target: < 1 hour
Met
Failover Tests
Monthly
Target: Quarterly
Exceeded
Last DR Drill
28 days ago
Target: < 90 days
Met
DR Architecture
Primary Region: US-East
  • Azure AKS cluster (3 nodes)
  • Cosmos DB (multi-master)
  • Cognitive Search (standard tier)
DR Region: US-West
  • Hot standby AKS cluster
  • Geo-replicated storage
  • Traffic Manager failover
Tested Scenarios
ScenarioExpectedActualStatus
Single AZ failureAuto-heal <5min3 minutes
Region failureManual failover <4hr2.5 hours
Data corruptionPoint-in-time <1hr45 minutes

Production RAG with governed retrieval and compliance gates.

CoCo doesn't replace Veeva, SharePoint, or Workfront. It sits on top. The systems of record remain authoritative. CoCo indexes their content, understands relationships, and retrieves the right information—with every answer traceable back to its source.

Built with clear SLOs: sub-second retrieval, 99.9% uptime, full audit trail. Every response includes citations. Every query is logged. MLR gateway validates compliance before content surfaces to users.

Answers: "What's the fair balance for Abrysvo?"
Finds: "Find approved Comirnaty templates for US market."
Guides: "How do I submit to MLR?"
Engineer Solution

CoCo Service Health Dashboard

Real-time monitoring across all platform services. Target SLOs: p95 <300ms, 99.7% uptime, <0.1% error rate.

RAG Pipeline Healthy
Uptime
99.7%
p95 Latency
280ms
Error Rate
0.03%
MLR Gateway Healthy
Uptime
99.9%
p95 Latency
85ms
Error Rate
0.01%
Vector Search Healthy
Uptime
99.8%
p95 Latency
65ms
Error Rate
0.02%
Citation Engine Healthy
Uptime
99.95%
p95 Latency
35ms
Error Rate
0.01%
Teams Bot Healthy
Uptime
99.6%
p95 Latency
380ms
Error Rate
0.05%
Audit Logger Healthy
Uptime
99.99%
p95 Latency
15ms
Error Rate
0.001%
API Gateway Latency (24h)
020040060000:0006:0012:0018:0023:00
p50
p95
p99
Aggregate Metrics
Avg Uptime
99.82%
Avg Error Rate
0.02%
Total Throughput
34.4K
req/min
Active Incidents
0
services affected
Active Alerts
✓ All systems operational

Agentic onboarding: learn by doing, not by reading manuals.

Traditional onboarding fails at scale. You can't have 200 trainers for 1,000 new hires. CoCo turns onboarding into an agentic experience: new employees ask questions as they arise, get immediate answers with context, and become productive in days instead of months.

Trained 200+ users across content production, MLR review, and brand management. Created playbooks for common workflows. Established feedback loops to continuously improve retrieval quality.

Before

  • ✕ Week-long training
  • ✕ 200-page manual
  • ✕ 3-4 months to productive

After

  • ✓ Day-one access
  • ✓ Ask as you go
  • ✓ Productive in days
Enable Solution

New Hire Onboarding Journey

From first day to first approval in under a week — powered by CoCo's guided assistance.

Goal: Create compliant content for Paxlovid campaign within first week
Day 1
Morning
First Search
Day 1-2
Content Creation
Day 2-3
MLR Submission
Day 3-4
First Approval
Day 5-7
EMOTIONAL JOURNEY
UncertainRelievedConfidentAccomplishedProud
TOUCHPOINTS
Teams Welcome
CoCo Introduction
CoCo Search
Template Library
CoCo Q&A
Claims Library
CoCo Process Guide
MLR System
MLR Notification
CoCo Celebration
PAIN POINTS
Overwhelmed by new systems
Learning brand specifics
Compliance uncertainty
Process complexity
OPPORTUNITIES
Immediate access to CoCo
Contextual suggestions
Real-time compliance hints
MLR pre-flight check
Success tracking dashboard

The Result

Publicis onboarded nearly 1,000 staff in time for peak vaccine season. MLR cycles dropped from 42 days to 14 days. Content throughput tripled. New hires were productive in days because CoCo was always there to answer questions, find content, and guide them through unfamiliar processes.

CoCo became the default way to work—not because it was mandated, but because it was faster. Today it serves as Pfizer's vaccines content backbone, with expansion planned for additional therapeutic areas.

42 days14 daysMLR Cycle Time
272/mo816/moAsset Throughput
25 min<1 minContent Search
Impact

Annual Value Creation — $2.08M

Breakdown of cost savings and productivity gains from CoCo deployment.

$0M$50M$100M$150MSearch Savings: +$78M$78MMLR Reduction: +$62.4M+$62.4MReduced Rework: +$45.6M+$45.6MFaster Onboarding: +$22.2M+$22.2MTotal Savings: +$208.2M$208.2MSearch SavingsMLR ReductionReduced ReworkFaster OnboardingTotal Savings
Total
Increase
Decrease

Total Cost of Ownership Analysis

Investment breakdown vs. realized savings — 96.5% ROI in Year 1.

Annual Investment: $1.06M
Azure Infrastructure$222K
OpenAI API Costs$98K
Cognitive Search$54K
Development Team$540K
Maintenance & Support$144K
Annual Savings: $2.08M
Search Time Savings$780K
MLR Cycle Reduction$624K
Reduced Rework$456K
Faster Onboarding$222K
Net Annual Benefit
$1.02M
Payback Period
6 months
Year 1 ROI
96.5%
3-Year ROI
490%

Lessons Learned

What Worked Well
  • Embedding with users for 2 weeks before coding — shadowing revealed the real bottlenecks
  • Graph-RAG for relationship mapping — dramatically improved cross-brand content discovery
  • Teams integration — zero friction adoption because CoCo met users where they worked
What We'd Do Differently
  • Started with smaller pilot (50 users) before scaling — initial feedback loop was too slow
  • Invested earlier in prompt engineering playbooks — inconsistent prompts caused retrieval variance
  • Built synthetic test datasets sooner — production testing delayed the feedback cycle

"The biggest ROI came not from the AI itself, but from finally having a single source of truth. CoCo forced us to clean up 20+ fragmented systems into one governed knowledge layer."