AI Book Companion - Principle-Based Management at Scale

How we built a multi-agent AI companion that brings Principle-Based Management to life for book readers and Koch Industries employees alike

When Koch Industries set out to bring Principle-Based Management (PBM) to a broad audience—both the reading public and Koch employees—they engaged Tribe AI to build an intelligent Book Companion. This AI-powered platform helps readers and employees alike learn, internalize, and apply PBM principles from the book by Chase and Charles Koch.

Analytiq Hub served as a subcontractor on the engagement, providing hands-on technical leadership and architecture for a production-grade multi-agent system with enterprise authentication, per-user memory, observability, and a progressive disclosure UX—all built to handle 8,000 to 100,000 users per month.

The Challenge

Koch Industries needed to make Principle-Based Management accessible and actionable for two distinct audiences: the general public reading the book, and Koch employees applying PBM in their daily work—with projected usage ranging from 8,000 to 100,000 users per month. Traditional approaches couldn’t scale to provide personalized, context-aware guidance on applying PBM principles to real-world scenarios. The solution required enterprise-grade security and compliance, seamless authentication supporting both public users and Koch employees via SSO, robust audit logging and data retention policies, and the ability to distinguish between principles and personal experiences in AI-generated responses—all while maintaining the nuance and depth that PBM education demands.

The Solution

Analytiq Hub helped architect a multi-agent AI Book Companion with three specialized agents—Learn, Prepare, and Reflect—coordinated by an orchestration layer that routes user interactions to the appropriate agent. The system features per-user memory for personalized learning journeys, enterprise authentication, comprehensive observability, and a feedback mechanism that drives continuous improvement. Deployed on AWS infrastructure, the platform has been validated through user testing and is advancing toward a broad public launch.

Key Capabilities That Drive Results

The AI Book Companion delivers enterprise-ready features optimized for organizational learning at scale:

  • Multi-Agent Architecture: Three specialized agents (Learn, Prepare, Reflect) with a router/orchestrator that intelligently directs user interactions to the right agent based on intent and context.
  • Per-User Memory System: MEM0-powered memory with PGVector and AWS Bedrock enables personalized learning journeys that remember each user’s progress, preferences, and areas of focus.
  • Enterprise Authentication: AWS Cognito and Auth0 integration with SSO support for Koch employees, including guest-to-authenticated user transitions and robust session management.
  • Observability & Tracing: Langfuse integration provides high-level agent traces, prompt monitoring, and memory usage tracking, complemented by CloudWatch for low-level infrastructure logs.
  • Progressive Disclosure UX: React frontend with a thoughtfully designed progressive disclosure pattern that guides users through increasingly deeper engagement with PBM content.

Platform Architecture That Scales

The solution leverages enterprise-grade infrastructure to support tens of thousands of users—both public readers and Koch employees:

Intelligent Multi-Agent Orchestration

Purpose-Built Agent Pipeline: The system routes user interactions through specialized agents that understand the nuances of Principle-Based Management:

  • Learn Agent: Guides users through PBM concepts, connecting principles to practical applications with source-grounded responses.
  • Prepare Agent: Helps users plan how to apply PBM principles to upcoming business scenarios and decisions.
  • Reflect Agent: Facilitates post-action reflection, helping users evaluate outcomes through a PBM lens.
  • Router/Orchestrator: Analyzes user intent and context to seamlessly direct conversations to the appropriate agent.

Enterprise-Ready Technology Stack

Built for Koch-Scale Deployment: The architecture ensures reliability, security, and compliance:

  • AWS Bedrock: Claude models power the AI agents, with ongoing model evaluation for optimal performance.
  • MEM0 + PGVector: Vector database backend for per-user memory storage and retrieval, enabling personalized and contextual interactions.
  • AWS Cognito/Auth0: Authentication supporting both public sign-up and enterprise SSO for Koch employees, with an admin portal for managing up to 100,000 monthly active users.
  • Langfuse: Production observability platform for tracing agent interactions, monitoring prompt performance, and debugging at scale.
  • React Frontend: Progressive disclosure UX that makes complex PBM content approachable and engaging.

Production-Grade Infrastructure

Reliable and Cost-Effective Operations: The infrastructure strategy balances performance with cost efficiency:

  • AWS/Kubernetes: Cluster deployments with VPC configurations optimized for security requirements and cross-account authentication.
  • Per-User Cost Modeling: Detailed cost estimation frameworks enabling Koch to plan infrastructure spend as usage scales.
  • Provisioned Capacity: Load testing and scalability planning to handle up to 100,000 monthly users during peak periods.
  • Multi-Tier Logging: Structured approach separating infrastructure logs (CloudWatch) from application-level traces (Langfuse) for rapid troubleshooting.

Analytiq Hub’s Contributions

As a subcontractor providing hands-on technical leadership, Analytiq Hub drove the project across multiple dimensions:

Infrastructure & DevOps: Led Langfuse deployment for observability, coordinated AWS/Kubernetes cluster deployments and VPC configurations, resolved cross-account authentication issues, and created per-user cost models for infrastructure planning.

Evaluation & Quality: Designed golden Q&A datasets to steer agent behavior toward principles-based responses, synthesized tester feedback into actionable improvement areas that drove feature prioritization, and established the multi-tier logging strategy.

Authentication & User Management: Architected the Auth0/Cognito integration including guest-to-authenticated user transitions and SSO requirements, identified the need for an admin portal at scale, and presented the memory system architecture to Koch leadership.

Memory System: Tested and debugged the MEM0 memory integration, ran debugging sessions for memory storage and retrieval, and helped navigate MEM0 vs. custom Postgres architectural decisions.

Engineering Leadership: Facilitated daily stand-ups, removed blockers for the engineering team, participated in onsite demos for Koch leadership, translated client feedback into technical requirements, and mentored team members through architectural decisions.

Security & Compliance: Tracked privacy policy requirements, log retention policies, audit logging, and account deletion workflows to meet Koch’s enterprise compliance standards.

Transformational Business Results

Validated at Scale: User testing confirmed that the multi-agent architecture, memory system, and authentication flows meet enterprise requirements—providing a solid foundation for broad rollout.

Scale-Ready Architecture: Platform designed and tested to support 8,000 to 100,000 monthly active users—both public readers and Koch employees—with provisioned capacity, dual authentication paths, and comprehensive admin tooling.

Accelerated Time to Production: By providing experienced technical leadership, Analytiq Hub helped the team move rapidly from concept to a production-grade platform, unblocking engineers and resolving complex infrastructure challenges along the way.

Enterprise Trust Established: Delivering production-grade infrastructure, security compliance, and hands-on technical leadership earned confidence for an ongoing partnership in scaling PBM education across Koch’s global organization.