letta-server/fern/pages/tutorials/draft/shared-memory-4-enterprise.mdx

---
title: "Shared Memory Part 4: Enterprise Multi-Team System"
subtitle: Build a complete enterprise system with departments and hierarchies
slug: cookbooks/shared-memory-enterprise
---

This tutorial demonstrates how to build a complete enterprise multi-agent system with multiple departments, hierarchical block architecture, and cross-team coordination.

## What You'll Learn

- Creating hierarchical block architectures (company → department → team)
- Implementing department-level isolation
- Building cross-department coordination mechanisms
- Managing executive-level aggregation and oversight
- Scaling shared memory to 10+ agents with complex access patterns

## Prerequisites

```bash
pip install letta-client
```

Set your Letta API key:
```bash
export LETTA_API_KEY="your-api-key-here"
```

## Architecture Overview

This tutorial creates a complete enterprise system with:
- **3 Departments**: Sales, Engineering, HR
- **10 Agents**: 1 CEO, 3 Directors, 6 Workers
- **12 Memory Blocks**: Company-wide, department-specific, cross-department, executive, and private blocks

```
TechCorp AI Solutions
├── CEO (Executive Dashboard)
├── Sales Department
│   ├── Sales Director
│   ├── Sales Rep 1
│   └── Sales Rep 2
├── Engineering Department
│   ├── Engineering Director
│   ├── Engineer 1
│   └── Engineer 2
└── HR Department
    ├── HR Director
    └── HR Rep 1
```

## Part 1: Company-Wide Blocks

```python
"""
Tutorial 4: Enterprise Multi-Team System
========================================

Build a complete enterprise with departments and hierarchies.
"""

from letta import Letta

# Initialize client
client = Letta()

print("=" * 70)
print("TUTORIAL 4: Enterprise Multi-Team System")
print("=" * 70)
print()

# ============================================================================
# STEP 1: Create Company-Wide Blocks (Read-Only, All Agents)
# ============================================================================
print("STEP 1: Creating company-wide blocks (all agents, read-only)...\n")

company_mission = client.blocks.create(
    label="company_mission",
    description="Company mission, values, and vision. Read-only for all agents.",
    value="""
=== TECHCORP AI SOLUTIONS ===
Company Mission & Values

MISSION:
Building ethical AI systems that empower businesses and respect user privacy.

CORE VALUES:
1. Transparency - Clear about AI capabilities and limitations
2. Privacy - User data protection is paramount
3. Accuracy - Rigorous testing and validation
4. Collaboration - Cross-functional teamwork
5. Innovation - Pushing boundaries responsibly

VISION 2025:
Become the most trusted AI solutions provider for enterprise customers.

COMPANY GUIDELINES:
- Always prioritize ethical considerations
- Maintain open communication across departments
- Customer success is our success
- Continuous learning and improvement
""",
    limit=4000
)

print(f"✓ Created company_mission: {company_mission.id}")
print(f"  Access: ALL 10 agents (read-only)")
print()

company_policies = client.blocks.create(
    label="company_policies",
    description="Company-wide policies and procedures. Read-only for all agents.",
    value="""
=== COMPANY POLICIES ===

COMMUNICATION POLICY:
- Respond to customer inquiries within 24 hours
- Internal team communication: Slack preferred
- Cross-department coordination: Use shared project blocks

DATA SECURITY POLICY:
- Never share customer data across departments without consent
- HR data is strictly confidential (HR department only)
- Financial data requires executive approval to share

WORK HOURS:
- Core hours: 10 AM - 4 PM (all time zones)
- Flexible scheduling outside core hours
- Respect work-life balance

ESCALATION POLICY:
- Team member → Department Director → CEO
- Cross-department issues → Use cross_dept_projects block
- Urgent matters → CEO direct escalation
""",
    limit=5000
)

print(f"✓ Created company_policies: {company_policies.id}")
print(f"  Access: ALL 10 agents (read-only)")
print()

# ============================================================================
# STEP 2: Create Department-Specific Blocks (Read/Write, Department Only)
# ============================================================================
print("STEP 2: Creating department-specific blocks...\n")

sales_knowledge = client.blocks.create(
    label="sales_knowledge",
    description="Sales team knowledge base. Sales department only.",
    value="""
=== SALES KNOWLEDGE BASE ===
Last Updated: 2024-10-08

PRODUCT PRICING:
- Starter Plan: $99/month
- Professional Plan: $299/month
- Enterprise Plan: Custom pricing (contact sales)

SALES PLAYBOOK:
- Lead qualification: BANT framework (Budget, Authority, Need, Timeline)
- Demo best practices: Focus on customer pain points
- Closing techniques: Trial close, assumptive close

CUSTOMER ACCOUNTS:
- Total Active Customers: 247
- Pipeline Value: $2.3M
- Avg Deal Size: $15,000

COMPETITIVE INTELLIGENCE:
- Main competitors: CompetitorX, CompetitorY
- Our advantages: Better privacy features, faster implementation
- Price positioning: Premium but justified by ROI

SALES TARGETS Q4:
- Team Target: $1.5M in new revenue
- Rep 1 Target: $600K
- Rep 2 Target: $600K
- Director Target: Strategic accounts $300K
""",
    limit=10000
)

print(f"✓ Created sales_knowledge: {sales_knowledge.id}")
print(f"  Access: Sales Director, Sales Rep 1, Sales Rep 2 (read/write)")
print()

engineering_specs = client.blocks.create(
    label="engineering_specs",
    description="Engineering specifications and technical docs. Engineering department only.",
    value="""
=== ENGINEERING SPECIFICATIONS ===
Last Updated: 2024-10-08

CURRENT SPRINT (Oct 8-22):
- Story Points: 42
- Completed: 18
- In Progress: 12
- Blocked: 0

ACTIVE PROJECTS:
1. API v3.0 Migration
   Status: 60% complete
   Owner: Engineer 1
   Priority: HIGH

2. Performance Optimization
   Status: 30% complete
   Owner: Engineer 2
   Priority: MEDIUM

3. Security Audit Remediation
   Status: In Planning
   Owner: Engineering Director
   Priority: HIGH

TECHNICAL STACK:
- Backend: Python, FastAPI
- Frontend: React, TypeScript
- Database: PostgreSQL, Redis
- Infrastructure: AWS, Kubernetes

ENGINEERING STANDARDS:
- Code review required for all PRs
- 80% test coverage minimum
- Documentation required for new features
- Weekly architecture reviews

KNOWN ISSUES:
- API latency on large datasets (Engineer 1 investigating)
- Mobile app occasional crashes (Engineer 2 working on fix)
""",
    limit=12000
)

print(f"✓ Created engineering_specs: {engineering_specs.id}")
print(f"  Access: Engineering Director, Engineer 1, Engineer 2 (read/write)")
print()

hr_employee_data = client.blocks.create(
    label="hr_employee_data",
    description="HR employee data and records. HR department only - CONFIDENTIAL.",
    value="""
=== HR EMPLOYEE DATA ===
Last Updated: 2024-10-08
CONFIDENTIAL - HR DEPARTMENT ONLY

EMPLOYEE HEADCOUNT:
- Total Employees: 47
- Sales: 12
- Engineering: 18
- HR: 4
- Operations: 8
- Executive: 5

ACTIVE HR CASES:
1. Employee ID #2847 - Performance improvement plan
   Status: Week 2 of 6
   Handler: HR Rep 1

2. Employee ID #1923 - Parental leave request
   Status: Approved, starts Oct 15
   Handler: HR Director

RECRUITING:
- Open Positions: 5
  * Senior Engineer (2 openings)
  * Sales Rep (1 opening)
  * HR Coordinator (1 opening)
  * Product Manager (1 opening)
- Interviews This Week: 8 scheduled

BENEFITS & COMPENSATION:
- Annual review cycle: November
- Benefits enrollment: Open now through Oct 31
- 401k match: 5%

COMPANY CULTURE:
- Employee satisfaction: 4.2/5 (recent survey)
- Retention rate: 92%
- Diversity initiatives: On track
""",
    limit=10000
)

print(f"✓ Created hr_employee_data: {hr_employee_data.id}")
print(f"  Access: HR Director, HR Rep 1 ONLY (read/write)")
print(f"  ⚠️  CONFIDENTIAL - Other departments CANNOT access")
print()

# ============================================================================
# STEP 3: Create Cross-Department Coordination Block
# ============================================================================
print("STEP 3: Creating cross-department coordination block...\n")

cross_dept_projects = client.blocks.create(
    label="cross_dept_projects",
    description="Cross-department projects and coordination. All directors + CEO.",
    value="""
=== CROSS-DEPARTMENT PROJECTS ===
Last Updated: 2024-10-08

ACTIVE CROSS-DEPT PROJECTS:
----------------------------

PROJECT: Enterprise Customer Onboarding Improvement
Teams: Sales + Engineering
Status: Planning Phase
Priority: HIGH

Background:
- Sales is closing large enterprise deals
- Engineering needs to improve onboarding experience
- Customer feedback: Onboarding takes too long

Action Items:
- [ ] Sales: Provide customer feedback summary (Due: Oct 10)
- [ ] Engineering: Audit current onboarding flow (Due: Oct 12)
- [ ] Joint meeting scheduled: Oct 15

PROJECT: Recruiting Pipeline for Engineering
Teams: HR + Engineering
Status: Active
Priority: MEDIUM

Background:
- 2 senior engineer positions open
- Need technical interview process improvement

Action Items:
- [x] HR: Post job descriptions (Completed)
- [ ] Engineering: Review candidates (Ongoing)
- [ ] HR + Engineering: Interview coordination

COMPLETED PROJECTS:
----------------------------
- Q3 Sales Enablement Training (Sales + Engineering) - Sep 2024
""",
    limit=8000
)

print(f"✓ Created cross_dept_projects: {cross_dept_projects.id}")
print(f"  Access: CEO, Sales Director, Engineering Director, HR Director (read/write)")
print(f"  Purpose: Coordinate work across departments")
print()

# ============================================================================
# STEP 4: Create Executive Dashboard Block (CEO Only)
# ============================================================================
print("STEP 4: Creating executive dashboard block (CEO only)...\n")

executive_dashboard = client.blocks.create(
    label="executive_dashboard",
    description="Executive dashboard with company-wide metrics. CEO only.",
    value="""
=== EXECUTIVE DASHBOARD ===
Last Updated: 2024-10-08
CEO EYES ONLY

COMPANY HEALTH METRICS:
------------------------
Revenue (Q4 to date): $3.2M / $4.5M target (71%)
Customer Satisfaction: 4.3/5
Employee Satisfaction: 4.2/5
Churn Rate: 3% (target: <5%)

DEPARTMENT PERFORMANCE:
------------------------
Sales:
- Pipeline: $2.3M (healthy)
- Q4 Target: $1.5M new revenue
- On track: Yes

Engineering:
- Sprint velocity: Steady
- Critical projects: 2 on track, 1 needs attention
- Technical debt: Manageable

HR:
- Headcount: 47 (target: 50 by EOY)
- Open positions: 5
- Retention: 92% (excellent)

STRATEGIC PRIORITIES:
------------------------
1. Close 3 enterprise deals by EOQ (Sales + Engineering)
2. Complete API v3.0 migration (Engineering)
3. Hire 2 senior engineers (HR + Engineering)
4. Maintain >90% employee retention (HR)

RISK FACTORS:
------------------------
- Competition increasing in enterprise segment
- Engineering team at capacity
- Need to scale customer support
""",
    limit=8000
)

print(f"✓ Created executive_dashboard: {executive_dashboard.id}")
print(f"  Access: CEO ONLY (read/write)")
print(f"  ⚠️  Executive-level metrics not visible to departments")
print()

print("=" * 70)
print("BLOCK ARCHITECTURE CREATED")
print("=" * 70)
print("""
BLOCK HIERARCHY:
================

TIER 1: Company-Wide (Read-Only, All 10 agents)
  - company_mission
  - company_policies

TIER 2: Department-Specific (Read/Write, Department only)
  - sales_knowledge (Sales dept: 3 agents)
  - engineering_specs (Engineering dept: 3 agents)
  - hr_employee_data (HR dept: 2 agents)

TIER 3: Cross-Department (Read/Write, Directors + CEO)
  - cross_dept_projects (4 agents: CEO + 3 Directors)

TIER 4: Executive (Read/Write, CEO only)
  - executive_dashboard (1 agent: CEO)

Total Shared Blocks: 7
Private Blocks per Agent: 1 each (10 total)
""")
```

<Accordion title="Expected Output">
```
======================================================================
TUTORIAL 4: Enterprise Multi-Team System
======================================================================

STEP 1: Creating company-wide blocks (all agents, read-only)...

✓ Created company_mission: block-p1q2r3s4-t5u6-7890-pqrs-tu1234567890
  Access: ALL 10 agents (read-only)

✓ Created company_policies: block-q2r3s4t5-u6v7-8901-qrst-uv2345678901
  Access: ALL 10 agents (read-only)

STEP 2: Creating department-specific blocks...

✓ Created sales_knowledge: block-r3s4t5u6-v7w8-9012-rstu-vw3456789012
  Access: Sales Director, Sales Rep 1, Sales Rep 2 (read/write)

✓ Created engineering_specs: block-s4t5u6v7-w8x9-0123-stuv-wx4567890123
  Access: Engineering Director, Engineer 1, Engineer 2 (read/write)

✓ Created hr_employee_data: block-t5u6v7w8-x9y0-1234-tuvw-xy5678901234
  Access: HR Director, HR Rep 1 ONLY (read/write)
  ⚠️  CONFIDENTIAL - Other departments CANNOT access

STEP 3: Creating cross-department coordination block...

✓ Created cross_dept_projects: block-u6v7w8x9-y0z1-2345-uvwx-yz6789012345
  Access: CEO, Sales Director, Engineering Director, HR Director (read/write)
  Purpose: Coordinate work across departments

STEP 4: Creating executive dashboard block (CEO only)...

✓ Created executive_dashboard: block-v7w8x9y0-z1a2-3456-vwxy-za7890123456
  Access: CEO ONLY (read/write)
  ⚠️  Executive-level metrics not visible to departments

======================================================================
BLOCK ARCHITECTURE CREATED
======================================================================
...
```
</Accordion>

## Part 2: Agent Creation

### Executive Layer

```python
# ============================================================================
# STEP 5: Create CEO Agent
# ============================================================================
print("\nSTEP 5: Creating CEO Agent (access to all cross-dept and executive blocks)...\n")

ceo_notes = client.blocks.create(
    label="ceo_executive_notes",
    description="Private CEO notes and strategic planning.",
    value="""
=== CEO EXECUTIVE NOTES ===

STRATEGIC FOCUS Q4:
- Enterprise market expansion
- Team scaling (controlled growth)
- Product-market fit validation

KEY RELATIONSHIPS:
- Board meeting: Monthly
- Investor updates: Quarterly
- All-hands: Bi-weekly

DECISION LOG:
(None yet)
""",
    limit=5000
)

ceo = client.agents.create(
    name="CEO",
    model="openai/gpt-4o",
    embedding="openai/text-embedding-3-small",
    memory_blocks=[
        {
            "label": "persona",
            "value": """I am the CEO of TechCorp AI Solutions.

My role:
- Set company vision and strategy
- Monitor overall company health
- Coordinate cross-department initiatives
- Make executive decisions
- Support department directors

My access:
- Company-wide blocks: company_mission, company_policies (read-only)
- Cross-department: cross_dept_projects (read/write)
- Executive: executive_dashboard (read/write, private)
- CEO notes (private)

Note: I do NOT have direct access to department-specific blocks (sales_knowledge,
engineering_specs, hr_employee_data). I rely on directors to surface relevant
information through cross_dept_projects.

My style:
- Strategic and big-picture focused
- Empowering to directors
- Data-driven decision making
- Transparent communication
"""
        },
        {
            "label": "human",
            "value": "Name: Board/Stakeholder\nRole: Company oversight"
        }
    ],
    block_ids=[company_mission.id, company_policies.id, cross_dept_projects.id, executive_dashboard.id, ceo_notes.id],
    tools=["core_memory_append", "core_memory_replace"],
)

print(f"✓ Created CEO: {ceo.id}")
print(f"  Shared blocks: company_mission, company_policies, cross_dept_projects, executive_dashboard")
print(f"  Private blocks: ceo_executive_notes")
print(f"  Note: NO direct access to department blocks")
print()

### Sales Department

print("STEP 6: Creating Sales Department agents...\n")

# Sales Director
sales_director_notes = client.blocks.create(
    label="sales_director_notes",
    description="Private sales director notes and strategy.",
    value="""
=== SALES DIRECTOR NOTES ===

STRATEGIC ACCOUNTS:
- 3 enterprise deals in pipeline
- Total value: $850K

TEAM MANAGEMENT:
- Rep 1: Strong performer, on track
- Rep 2: Ramping up, needs coaching

PRIORITIES:
- Close Acme Corp deal ($300K)
- Coordinate with Engineering on onboarding
""",
    limit=5000
)

sales_director = client.agents.create(
    name="Sales_Director",
    model="openai/gpt-4o",
    embedding="openai/text-embedding-3-small",
    memory_blocks=[
        {
            "label": "persona",
            "value": """I am the Sales Director at TechCorp AI Solutions.

My role:
- Lead sales team and strategy
- Manage strategic accounts
- Coordinate with other departments (Engineering, HR)
- Report to CEO on sales performance

My access:
- Company-wide: company_mission, company_policies (read-only)
- Department: sales_knowledge (read/write, shared with team)
- Cross-dept: cross_dept_projects (read/write, coordinate with other directors)
- Private: sales_director_notes

My style:
- Results-oriented and strategic
- Collaborative with Engineering
- Supportive of sales reps
- Customer-focused
"""
        },
        {
            "label": "human",
            "value": "Name: CEO or Sales Team\nRole: Sales leadership"
        }
    ],
    block_ids=[company_mission.id, company_policies.id, sales_knowledge.id, cross_dept_projects.id, sales_director_notes.id],
    tools=["core_memory_append", "core_memory_replace"],
)

print(f"✓ Created Sales Director: {sales_director.id}")
print()

# Sales Rep 1
sales_rep1_notes = client.blocks.create(
    label="sales_rep1_notes",
    description="Private sales rep 1 notes and leads.",
    value="=== SALES REP 1 NOTES ===\n\nActive Leads: 12\nQ4 Target: $600K\nClosed This Quarter: $420K\n",
    limit=4000
)

sales_rep1 = client.agents.create(
    name="Sales_Rep_1",
    model="openai/gpt-4o-mini",
    embedding="openai/text-embedding-3-small",
    memory_blocks=[
        {
            "label": "persona",
            "value": """I am Sales Rep 1 at TechCorp AI Solutions.

My role:
- Manage sales pipeline
- Close deals
- Coordinate with Sales Director
- Reference sales playbook and pricing

My access:
- Company-wide: company_mission, company_policies (read-only)
- Department: sales_knowledge (read/write, shared with sales team)
- Private: sales_rep1_notes

My style:
- Customer-focused and consultative
- Data-driven
- Team player
"""
        },
        {
            "label": "human",
            "value": "Name: Sales Director or Customer\nRole: Sales execution"
        }
    ],
    block_ids=[company_mission.id, company_policies.id, sales_knowledge.id, sales_rep1_notes.id],
    tools=["core_memory_append", "core_memory_replace"],
)

print(f"✓ Created Sales Rep 1: {sales_rep1.id}")
print()

# Sales Rep 2
sales_rep2_notes = client.blocks.create(
    label="sales_rep2_notes",
    description="Private sales rep 2 notes and leads.",
    value="=== SALES REP 2 NOTES ===\n\nActive Leads: 15\nQ4 Target: $600K\nClosed This Quarter: $380K\n",
    limit=4000
)

sales_rep2 = client.agents.create(
    name="Sales_Rep_2",
    model="openai/gpt-4o-mini",
    embedding="openai/text-embedding-3-small",
    memory_blocks=[
        {
            "label": "persona",
            "value": """I am Sales Rep 2 at TechCorp AI Solutions.

My role:
- Manage sales pipeline
- Close deals
- Coordinate with Sales Director
- Reference sales playbook and pricing

My access:
- Company-wide: company_mission, company_policies (read-only)
- Department: sales_knowledge (read/write, shared with sales team)
- Private: sales_rep2_notes

My style:
- Relationship-focused
- Strategic account management
- Collaborative
"""
        },
        {
            "label": "human",
            "value": "Name: Sales Director or Customer\nRole: Sales execution"
        }
    ],
    block_ids=[company_mission.id, company_policies.id, sales_knowledge.id, sales_rep2_notes.id],
    tools=["core_memory_append", "core_memory_replace"],
)

print(f"✓ Created Sales Rep 2: {sales_rep2.id}")
print()

### Engineering Department

print("STEP 7: Creating Engineering Department agents...\n")

# Engineering Director
eng_director_notes = client.blocks.create(
    label="eng_director_notes",
    description="Private engineering director notes and planning.",
    value="""
=== ENGINEERING DIRECTOR NOTES ===

TECHNICAL STRATEGY:
- API v3.0 migration: Top priority
- Security audit: Must complete Q4
- Performance optimization: Ongoing

TEAM CAPACITY:
- Need 2 senior engineers
- Current team at 85% capacity
- Considering contractors for peak load

CROSS-DEPT COORDINATION:
- Sales asking for faster onboarding
- HR coordinating on hiring
""",
    limit=5000
)

eng_director = client.agents.create(
    name="Engineering_Director",
    model="openai/gpt-4o",
    embedding="openai/text-embedding-3-small",
    memory_blocks=[
        {
            "label": "persona",
            "value": """I am the Engineering Director at TechCorp AI Solutions.

My role:
- Lead engineering team and technical strategy
- Manage engineering projects and sprints
- Coordinate with Sales and HR departments
- Report to CEO on technical progress

My access:
- Company-wide: company_mission, company_policies (read-only)
- Department: engineering_specs (read/write, shared with engineers)
- Cross-dept: cross_dept_projects (read/write, coordinate with other directors)
- Private: eng_director_notes

My style:
- Technical and strategic
- Quality-focused
- Collaborative with Sales (understand customer needs)
- Supportive of engineering team
"""
        },
        {
            "label": "human",
            "value": "Name: CEO or Engineering Team\nRole: Engineering leadership"
        }
    ],
    block_ids=[company_mission.id, company_policies.id, engineering_specs.id, cross_dept_projects.id, eng_director_notes.id],
    tools=["core_memory_append", "core_memory_replace"],
)

print(f"✓ Created Engineering Director: {eng_director.id}")
print()

# Engineer 1
engineer1_notes = client.blocks.create(
    label="engineer1_notes",
    description="Private engineer 1 notes and tasks.",
    value="=== ENGINEER 1 NOTES ===\n\nCurrent Project: API v3.0 Migration (60% complete)\nBlockers: None\nNext: Database schema updates\n",
    limit=4000
)

engineer1 = client.agents.create(
    name="Engineer_1",
    model="openai/gpt-4o-mini",
    embedding="openai/text-embedding-3-small",
    memory_blocks=[
        {
            "label": "persona",
            "value": """I am Engineer 1 at TechCorp AI Solutions.

My role:
- Develop and maintain software
- Work on assigned projects (currently: API v3.0 migration)
- Follow engineering standards
- Coordinate with Engineering Director

My access:
- Company-wide: company_mission, company_policies (read-only)
- Department: engineering_specs (read/write, shared with engineering team)
- Private: engineer1_notes

My style:
- Detail-oriented and thorough
- Quality-focused (testing, documentation)
- Collaborative with team
"""
        },
        {
            "label": "human",
            "value": "Name: Engineering Director or Team\nRole: Software development"
        }
    ],
    block_ids=[company_mission.id, company_policies.id, engineering_specs.id, engineer1_notes.id],
    tools=["core_memory_append", "core_memory_replace"],
)

print(f"✓ Created Engineer 1: {engineer1.id}")
print()

# Engineer 2
engineer2_notes = client.blocks.create(
    label="engineer2_notes",
    description="Private engineer 2 notes and tasks.",
    value="=== ENGINEER 2 NOTES ===\n\nCurrent Project: Performance Optimization (30% complete)\nBlockers: None\nNext: Query optimization for large datasets\n",
    limit=4000
)

engineer2 = client.agents.create(
    name="Engineer_2",
    model="openai/gpt-4o-mini",
    embedding="openai/text-embedding-3-small",
    memory_blocks=[
        {
            "label": "persona",
            "value": """I am Engineer 2 at TechCorp AI Solutions.

My role:
- Develop and maintain software
- Work on assigned projects (currently: Performance optimization)
- Follow engineering standards
- Coordinate with Engineering Director

My access:
- Company-wide: company_mission, company_policies (read-only)
- Department: engineering_specs (read/write, shared with engineering team)
- Private: engineer2_notes

My style:
- Performance-focused
- Data-driven optimization
- Pragmatic problem-solving
"""
        },
        {
            "label": "human",
            "value": "Name: Engineering Director or Team\nRole: Software development"
        }
    ],
    block_ids=[company_mission.id, company_policies.id, engineering_specs.id, engineer2_notes.id],
    tools=["core_memory_append", "core_memory_replace"],
)

print(f"✓ Created Engineer 2: {engineer2.id}")
print()

### HR Department

print("STEP 8: Creating HR Department agents...\n")

# HR Director
hr_director_notes = client.blocks.create(
    label="hr_director_notes",
    description="Private HR director notes and planning.",
    value="""
=== HR DIRECTOR NOTES ===

RECRUITING PRIORITIES:
- 2 Senior Engineers (urgent for Engineering)
- 1 Sales Rep (Sales scaling)
- 1 HR Coordinator (team growth)

EMPLOYEE RELATIONS:
- 1 performance improvement plan (ongoing)
- Annual review prep (November)
- Benefits enrollment (ends Oct 31)

CROSS-DEPT COORDINATION:
- Engineering: Hiring coordination
- Sales: New rep onboarding planning
""",
    limit=5000
)

hr_director = client.agents.create(
    name="HR_Director",
    model="openai/gpt-4o",
    embedding="openai/text-embedding-3-small",
    memory_blocks=[
        {
            "label": "persona",
            "value": """I am the HR Director at TechCorp AI Solutions.

My role:
- Lead HR strategy and operations
- Manage recruiting, employee relations, benefits
- Coordinate with other departments on hiring
- Maintain confidential employee data
- Report to CEO on HR metrics

My access:
- Company-wide: company_mission, company_policies (read-only)
- Department: hr_employee_data (read/write, CONFIDENTIAL, HR only)
- Cross-dept: cross_dept_projects (read/write, coordinate with other directors)
- Private: hr_director_notes

My style:
- People-focused and empathetic
- Confidentiality is paramount
- Strategic about talent and culture
- Collaborative with all departments
"""
        },
        {
            "label": "human",
            "value": "Name: CEO or HR Team\nRole: HR leadership"
        }
    ],
    block_ids=[company_mission.id, company_policies.id, hr_employee_data.id, cross_dept_projects.id, hr_director_notes.id],
    tools=["core_memory_append", "core_memory_replace"],
)

print(f"✓ Created HR Director: {hr_director.id}")
print()

# HR Rep 1
hr_rep1_notes = client.blocks.create(
    label="hr_rep1_notes",
    description="Private HR rep 1 notes and cases.",
    value="=== HR REP 1 NOTES ===\n\nActive Cases: 2\n- Performance improvement plan (Employee #2847)\n- Benefits questions (3 employees)\n",
    limit=4000
)

hr_rep1 = client.agents.create(
    name="HR_Rep_1",
    model="openai/gpt-4o-mini",
    embedding="openai/text-embedding-3-small",
    memory_blocks=[
        {
            "label": "persona",
            "value": """I am HR Rep 1 at TechCorp AI Solutions.

My role:
- Handle employee relations cases
- Support recruiting efforts
- Manage benefits administration
- Maintain confidentiality
- Coordinate with HR Director

My access:
- Company-wide: company_mission, company_policies (read-only)
- Department: hr_employee_data (read/write, CONFIDENTIAL, HR only)
- Private: hr_rep1_notes

My style:
- Empathetic and supportive
- Confidentiality is critical
- Detail-oriented with cases
- Helpful and responsive
"""
        },
        {
            "label": "human",
            "value": "Name: HR Director or Employees\nRole: HR operations"
        }
    ],
    block_ids=[company_mission.id, company_policies.id, hr_employee_data.id, hr_rep1_notes.id],
    tools=["core_memory_append", "core_memory_replace"],
)

print(f"✓ Created HR Rep 1: {hr_rep1.id}")
print()

print("=" * 70)
print("ENTERPRISE ORGANIZATION CREATED - 10 AGENTS")
print("=" * 70)
print("""
ORGANIZATION STRUCTURE:
=======================

CEO (1 agent)
  └── Access: company_mission, company_policies, cross_dept_projects, executive_dashboard

Sales Department (3 agents)
  ├── Sales Director
  │   └── Access: company blocks + sales_knowledge + cross_dept_projects
  ├── Sales Rep 1
  │   └── Access: company blocks + sales_knowledge
  └── Sales Rep 2
      └── Access: company blocks + sales_knowledge

Engineering Department (3 agents)
  ├── Engineering Director
  │   └── Access: company blocks + engineering_specs + cross_dept_projects
  ├── Engineer 1
  │   └── Access: company blocks + engineering_specs
  └── Engineer 2
      └── Access: company blocks + engineering_specs

HR Department (2 agents)
  ├── HR Director
  │   └── Access: company blocks + hr_employee_data + cross_dept_projects
  └── HR Rep 1
      └── Access: company blocks + hr_employee_data

BLOCK ACCESS SUMMARY:
=====================
- 10 agents total
- 7 shared blocks
- 10 private blocks (1 per agent)
- Department isolation enforced
- Cross-department coordination via cross_dept_projects
- Executive oversight via executive_dashboard
""")```

## Part 3: Message Flow Scenarios

### Scenario 1: Company-Wide Policy Check (Vertical Consistency)

```python
print("\n" + "=" * 70)
print("SCENARIO 1: Company-Wide Policy Check Across All Departments")
print("=" * 70)

# Test that all departments see the same company mission
print("\nTesting that Sales Rep, Engineer, and HR Rep all see same company mission...\n")

print("User → Sales Rep 1: 'What is our company mission?'\n")
response_sales = client.agents.messages.create(
    agent_id=sales_rep1.id,
    messages=[{
        "role": "user",
        "content": "What is our company's mission?"
    }]
)

for msg in response_sales.messages:
    if msg.message_type == "assistant_message":
        print(f"Sales Rep 1: {msg.content}\n")

print("User → Engineer 1: 'What is our company mission?'\n")
response_eng = client.agents.messages.create(
    agent_id=engineer1.id,
    messages=[{
        "role": "user",
        "content": "What is our company's mission?"
    }]
)

for msg in response_eng.messages:
    if msg.message_type == "assistant_message":
        print(f"Engineer 1: {msg.content}\n")

print("User → HR Rep 1: 'What is our company mission?'\n")
response_hr = client.agents.messages.create(
    agent_id=hr_rep1.id,
    messages=[{
        "role": "user",
        "content": "What is our company's mission?"
    }]
)

for msg in response_hr.messages:
    if msg.message_type == "assistant_message":
        print(f"HR Rep 1: {msg.content}\n")

print("Analysis:")
print("- All three agents from different departments accessed company_mission")
print("- All provided identical mission statement")
print("- Demonstrates vertical consistency across entire organization")
print("- Company-wide blocks ensure unified messaging")
```

<Accordion title="Expected Output">
```
======================================================================
SCENARIO 1: Company-Wide Policy Check Across All Departments
======================================================================

Testing that Sales Rep, Engineer, and HR Rep all see same company mission...

User → Sales Rep 1: 'What is our company mission?'

Sales Rep 1: Our company mission at TechCorp AI Solutions is "Building ethical AI
systems that empower businesses and respect user privacy." We focus on transparency,
privacy, accuracy, collaboration, and innovation as our core values.

User → Engineer 1: 'What is our company mission?'

Engineer 1: TechCorp's mission is to build ethical AI systems that empower businesses
while respecting user privacy. Our core values are transparency, privacy, accuracy,
collaboration, and innovation.

User → HR Rep 1: 'What is our company mission?'

HR Rep 1: Our mission is building ethical AI systems that empower businesses and
respect user privacy. We prioritize transparency, privacy, accuracy, collaboration,
and innovation in everything we do.

Analysis:
- All three agents from different departments accessed company_mission
- All provided identical mission statement
- Demonstrates vertical consistency across entire organization
- Company-wide blocks ensure unified messaging
```
</Accordion>

### Scenario 2: Department-Isolated Knowledge

```python
print("\n" + "=" * 70)
print("SCENARIO 2: Department Isolation - Sales Cannot Access Engineering Data")
print("=" * 70)
print("\nUser → Sales Rep 1: 'What engineering projects are currently in progress?'\n")

response = client.agents.messages.create(
    agent_id=sales_rep1.id,
    messages=[{
        "role": "user",
        "content": "I need to know what engineering projects are currently in progress. What's the status of the API v3.0 migration?"
    }]
)

for msg in response.messages:
    if msg.message_type == "assistant_message":
        print(f"Sales Rep 1: {msg.content}\n")

print("Analysis:")
print("- Sales Rep 1 does NOT have access to engineering_specs")
print("- Cannot answer specific engineering project questions")
print("- Should suggest asking Engineering Director or using cross_dept_projects")
print("- Demonstrates department isolation working correctly")
```

<Accordion title="Expected Output">
```
======================================================================
SCENARIO 2: Department Isolation - Sales Cannot Access Engineering Data
======================================================================

User → Sales Rep 1: 'What engineering projects are currently in progress?'

Sales Rep 1: I don't have direct access to Engineering's project details or the
engineering_specs block. For specific information about the API v3.0 migration status,
you should:

1. Ask the Engineering Director directly
2. Check the cross_dept_projects block if there's a related cross-department initiative
3. Our Sales Director can coordinate with Engineering leadership

I can tell you about our sales pipeline and customer needs, which might help inform
engineering priorities if needed. Would that be helpful?

Analysis:
- Sales Rep 1 does NOT have access to engineering_specs
- Cannot answer specific engineering project questions
- Should suggest asking Engineering Director or using cross_dept_projects
- Demonstrates department isolation working correctly
```
</Accordion>

<Warning>
Department isolation is critical for security and information compartmentalization. Agents correctly recognize their access boundaries and direct users to appropriate channels.
</Warning>

### Scenario 3: Cross-Department Coordination (Directors)

```python
print("\n" + "=" * 70)
print("SCENARIO 3: Sales Director Coordinates with Engineering via Cross-Dept Block")
print("=" * 70)
print("\nUser → Sales Director: 'Update the cross-department projects: Sales is hearing from customers that onboarding takes too long. We need Engineering's help to streamline it.'\n")

response = client.agents.messages.create(
    agent_id=sales_director.id,
    messages=[{
        "role": "user",
        "content": "We're getting feedback from customers that the onboarding process takes too long. This is affecting deal closures. Can you update the cross-department projects block to coordinate with Engineering on improving the onboarding experience?"
    }]
)

for msg in response.messages:
    if msg.message_type == "assistant_message":
        print(f"Sales Director: {msg.content}\n")

# Check updated cross_dept_projects
updated_cross_dept = client.blocks.retrieve(block_id=cross_dept_projects.id)
print("Updated cross_dept_projects block:")
print("-" * 70)
print(updated_cross_dept.value[:800] + "...")
print("-" * 70)

print("\nAnalysis:")
print("- Sales Director has access to cross_dept_projects")
print("- Updated block with customer feedback and action items")
print("- Engineering Director will see this update")
print("- Demonstrates director-level cross-department coordination")
```

<Accordion title="Expected Output">
```
======================================================================
SCENARIO 3: Sales Director Coordinates with Engineering via Cross-Dept Block
======================================================================

User → Sales Director: 'Update the cross-department projects...'

Sales Director: I've updated the cross_dept_projects block with this critical
customer feedback. Here's what I've documented:

**Added to Enterprise Customer Onboarding Improvement Project:**
- Priority escalated to HIGH (affecting deal closures)
- Customer feedback: Onboarding takes too long
- Impact: Customers hesitating on large deals
- Sales action item: Compile specific customer feedback with timeline expectations

The Engineering Director will see this update and can respond with their technical
assessment. This is exactly the kind of cross-functional coordination we need to
maintain our competitive edge.

Updated cross_dept_projects block:
----------------------------------------------------------------------
=== CROSS-DEPARTMENT PROJECTS ===
Last Updated: 2024-10-08 14:45 PM

ACTIVE CROSS-DEPT PROJECTS:
----------------------------

PROJECT: Enterprise Customer Onboarding Improvement
Teams: Sales + Engineering
Status: Planning Phase → ESCALATED
Priority: HIGH → CRITICAL

Background:
- Sales is closing large enterprise deals
- Engineering needs to improve onboarding experience
- Customer feedback: Onboarding takes too long (affecting deal closures)
- Sales reporting: Multiple customers citing onboarding as concern

Action Items:
- [x] Sales: Provide customer feedback summary (Sales Director - Completed)
- [ ] Engineering: Audit current onboarding flow (Due: Oct 12)
...
----------------------------------------------------------------------

Analysis:
- Sales Director has access to cross_dept_projects
- Updated block with customer feedback and action items
- Engineering Director will see this update
- Demonstrates director-level cross-department coordination
```
</Accordion>

### Scenario 4: Engineering Director Sees Sales Update

```python
print("\n" + "=" * 70)
print("SCENARIO 4: Engineering Director Responds to Sales Request")
print("=" * 70)
print("\nUser → Engineering Director: 'Check the cross-department projects block. Any updates from other teams?'\n")

response = client.agents.messages.create(
    agent_id=eng_director.id,
    messages=[{
        "role": "user",
        "content": "Check the cross-department projects block for any updates from other teams that need our attention."
    }]
)

for msg in response.messages:
    if msg.message_type == "assistant_message":
        print(f"Engineering Director: {msg.content}\n")

print("Analysis:")
print("- Engineering Director read cross_dept_projects")
print("- Saw Sales Director's escalation")
print("- Can now respond and coordinate internally")
print("- Demonstrates asynchronous cross-department communication via shared blocks")
```

<Accordion title="Expected Output">
```
======================================================================
SCENARIO 4: Engineering Director Responds to Sales Request
======================================================================

User → Engineering Director: 'Check the cross-department projects block...'

Engineering Director: I see an important escalation from the Sales Director:

**Enterprise Customer Onboarding Improvement - Now CRITICAL Priority**

The Sales team is reporting that onboarding delays are affecting deal closures. This
is serious and needs immediate attention.

Here's my response:

1. **Immediate Action**: I'll have Engineer 1 conduct an onboarding flow audit this week
2. **Technical Assessment**: We'll identify bottlenecks (likely account provisioning and initial setup)
3. **Timeline**: Preliminary findings by Oct 12, implementation plan by Oct 15
4. **Resource Allocation**: This takes priority over non-critical performance optimization work

I'll update the cross_dept_projects block with our technical assessment once Engineer 1
completes the audit. Let me also update the engineering_specs block to reflect this
new priority for my team.

This is excellent cross-department coordination - exactly what we need to stay customer-focused.

Analysis:
- Engineering Director read cross_dept_projects
- Saw Sales Director's escalation
- Can now respond and coordinate internally
- Demonstrates asynchronous cross-department communication via shared blocks
```
</Accordion>

<Tip>
The `cross_dept_projects` block acts as an **asynchronous coordination channel** between directors, eliminating the need for explicit messaging or meetings for routine coordination.
</Tip>

### Scenario 5: HR Data Confidentiality

```python
print("\n" + "=" * 70)
print("SCENARIO 5: Testing HR Data Confidentiality")
print("=" * 70)
print("\nUser → Sales Director: 'Who is currently on a performance improvement plan?'\n")

response = client.agents.messages.create(
    agent_id=sales_director.id,
    messages=[{
        "role": "user",
        "content": "I heard someone in the company is on a performance improvement plan. Who is it and what's the situation?"
    }]
)

for msg in response.messages:
    if msg.message_type == "assistant_message":
        print(f"Sales Director: {msg.content}\n")

print("Analysis:")
print("- Sales Director does NOT have access to hr_employee_data")
print("- Cannot access confidential HR information")
print("- Should redirect to HR Director")
print("- Demonstrates HR confidentiality protection")
```

<Accordion title="Expected Output">
```
======================================================================
SCENARIO 5: Testing HR Data Confidentiality
======================================================================

User → Sales Director: 'Who is currently on a performance improvement plan?'

Sales Director: I don't have access to confidential HR employee data, including
information about performance improvement plans. That information is restricted to
the HR department only, as per our company policies on data security and employee
privacy.

If you have a legitimate need to know about a specific employee situation (for example,
if it affects your team), you should:
1. Contact the HR Director directly
2. They'll assess whether sharing that information is appropriate
3. Any discussions will be handled confidentially

Is there a specific business concern I can help you with regarding team performance?

Analysis:
- Sales Director does NOT have access to hr_employee_data
- Cannot access confidential HR information
- Should redirect to HR Director
- Demonstrates HR confidentiality protection
```
</Accordion>

### Scenario 6: CEO Strategic Overview

```python
print("\n" + "=" * 70)
print("SCENARIO 6: CEO Reviews Company Status")
print("=" * 70)
print("\nUser → CEO: 'Give me a company-wide status update. What's happening across all departments?'\n")

response = client.agents.messages.create(
    agent_id=ceo.id,
    messages=[{
        "role": "user",
        "content": "I need a comprehensive company update. What's the status across Sales, Engineering, and HR? Any cross-department issues I should be aware of?"
    }]
)

for msg in response.messages:
    if msg.message_type == "assistant_message":
        print(f"CEO: {msg.content}\n")

print("Analysis:")
print("- CEO accessed executive_dashboard (company-wide metrics)")
print("- CEO accessed cross_dept_projects (coordination status)")
print("- CEO does NOT have direct access to department-specific blocks")
print("- CEO provides high-level overview based on aggregated data")
print("- Demonstrates executive oversight without micromanagement")
```

<Accordion title="Expected Output">
```
======================================================================
SCENARIO 6: CEO Reviews Company Status
======================================================================

User → CEO: 'Give me a company-wide status update...'

CEO: Here's the company-wide status based on my executive dashboard and cross-department
coordination:

**OVERALL COMPANY HEALTH:**
- Revenue: $3.2M / $4.5M Q4 target (71% - on track)
- Customer Satisfaction: 4.3/5 (strong)
- Employee Satisfaction: 4.2/5 (healthy)
- Churn Rate: 3% (below 5% target ✓)

**DEPARTMENT STATUS:**

*Sales*
✓ Pipeline: $2.3M (healthy)
✓ On track for $1.5M Q4 target
⚠️ Issue: Customer feedback on slow onboarding (being addressed)

*Engineering*
✓ Sprint velocity steady
✓ API v3.0 migration: 60% complete
⚠️ Team at capacity - coordinating with HR on hiring

*HR*
✓ Employee retention: 92% (excellent)
⚠️ 5 open positions (2 senior engineers critical for Engineering)
✓ Recruiting pipeline active

**CROSS-DEPARTMENT COORDINATION:**
Active: Onboarding improvement (Sales + Engineering) - escalated to CRITICAL priority
Active: Engineering hiring (HR + Engineering) - ongoing

**STRATEGIC ASSESSMENT:**
We're generally healthy, but the onboarding issue needs immediate attention as it's
affecting deal closures. The fact that Sales and Engineering are already coordinating
through cross_dept_projects is exactly the collaboration we need.

Recommendation: I'll monitor the onboarding improvement project closely. If they need
resources or executive intervention, I'm ready to support.

Analysis:
- CEO accessed executive_dashboard (company-wide metrics)
- CEO accessed cross_dept_projects (coordination status)
- CEO does NOT have direct access to department-specific blocks
- CEO provides high-level overview based on aggregated data
- Demonstrates executive oversight without micromanagement
```
</Accordion>

<Note>
The CEO has **strategic oversight** without micromanaging. The executive_dashboard provides aggregated metrics, while cross_dept_projects shows coordination, but the CEO doesn't see day-to-day department operations.
</Note>

### Scenario 7: Hierarchical Reporting (Bottom-Up)

```python
print("\n" + "=" * 70)
print("SCENARIO 7: Sales Rep Reports to Director (Bottom-Up Communication)")
print("=" * 70)
print("\nUser → Sales Rep 1: 'Log that you closed the Acme Corp deal for $300K in the sales knowledge base'\n")

response = client.agents.messages.create(
    agent_id=sales_rep1.id,
    messages=[{
        "role": "user",
        "content": "Great news! I just closed the Acme Corp deal for $300K. Update the sales knowledge base with this win and update my personal Q4 numbers."
    }]
)

for msg in response.messages:
    if msg.message_type == "assistant_message":
        print(f"Sales Rep 1: {msg.content}\n")

# Check updated sales_knowledge
updated_sales = client.blocks.retrieve(block_id=sales_knowledge.id)
print("Updated sales_knowledge block (excerpt):")
print("-" * 70)
print(updated_sales.value[:600] + "...")
print("-" * 70)

print("\nNow Sales Director checks the shared sales_knowledge...")
print("User → Sales Director: 'Any wins from the team today?'\n")

response = client.agents.messages.create(
    agent_id=sales_director.id,
    messages=[{
        "role": "user",
        "content": "Check the sales knowledge base - any wins from the team today?"
    }]
)

for msg in response.messages:
    if msg.message_type == "assistant_message":
        print(f"Sales Director: {msg.content}\n")

print("Analysis:")
print("- Sales Rep 1 updated shared sales_knowledge block")
print("- Sales Director immediately sees the update")
print("- Demonstrates bottom-up information flow within department")
print("- Director has visibility into team performance without explicit reporting")
```

<Accordion title="Expected Output">
```
======================================================================
SCENARIO 7: Sales Rep Reports to Director (Bottom-Up Communication)
======================================================================

User → Sales Rep 1: 'Log that you closed the Acme Corp deal for $300K...'

Sales Rep 1: Excellent! I've updated the sales knowledge base with the Acme Corp win:

**Deal Details Logged:**
- Customer: Acme Corp
- Deal Size: $300K
- Close Date: Oct 8, 2024
- Plan: Enterprise (custom pricing)
- Rep: Sales Rep 1

**Updated Q4 Numbers:**
- My Q4 closed: $420K → $720K
- Target: $600K
- Achievement: 120% of target ✓

This is a significant win! I've also updated my personal notes with next steps for
Acme's onboarding.

Updated sales_knowledge block (excerpt):
----------------------------------------------------------------------
=== SALES KNOWLEDGE BASE ===
Last Updated: 2024-10-08 15:30 PM

RECENT WINS:
- Acme Corp: $300K Enterprise deal (Sales Rep 1) - Oct 8, 2024

CUSTOMER ACCOUNTS:
- Total Active Customers: 248 (was 247)
- Pipeline Value: $2.0M (was $2.3M, Acme converted)
- Avg Deal Size: $15,100

SALES TARGETS Q4:
- Team Target: $1.5M in new revenue
- Rep 1 Target: $600K → Actual: $720K (120% ✓)
...
----------------------------------------------------------------------

Now Sales Director checks the shared sales_knowledge...
User → Sales Director: 'Any wins from the team today?'

Sales Director: Yes! Great news from the team:

**🎉 Major Win: Acme Corp - $300K Enterprise Deal**
Closed by: Sales Rep 1
Date: Today (Oct 8)

This is fantastic - Sales Rep 1 has now hit 120% of their Q4 target ($720K / $600K).
This was one of our strategic accounts, and the $300K deal size is well above our
average.

**Team Status:**
- Rep 1: $720K closed (120% of target) - exceeding expectations
- Rep 2: $380K closed (63% of target) - on track
- Team total: $1.1M / $1.5M (73%)

We're in great shape for Q4. I should recognize Rep 1's performance and make sure
we have a smooth onboarding for Acme.

Analysis:
- Sales Rep 1 updated shared sales_knowledge block
- Sales Director immediately sees the update
- Demonstrates bottom-up information flow within department
- Director has visibility into team performance without explicit reporting
```
</Accordion>

### Scenario 8: Information Cascade (Top-Down)

```python
print("\n" + "=" * 70)
print("SCENARIO 8: CEO Decision Cascades Down")
print("=" * 70)
print("\nUser → CEO: 'The board wants us to prioritize the enterprise segment. Update the strategy and make sure all departments know.'\n")

response = client.agents.messages.create(
    agent_id=ceo.id,
    messages=[{
        "role": "user",
        "content": "The board meeting went well. Key decision: We're doubling down on the enterprise segment. Update the executive dashboard and coordinate with department directors via the cross-department projects block."
    }]
)

for msg in response.messages:
    if msg.message_type == "assistant_message":
        print(f"CEO: {msg.content}\n")

print("Now check if Directors see the update...")
print("\nUser → Sales Director: 'Any strategic updates from leadership?'\n")

response = client.agents.messages.create(
    agent_id=sales_director.id,
    messages=[{
        "role": "user",
        "content": "Check the cross-department projects block - any strategic updates from the CEO?"
    }]
)

for msg in response.messages:
    if msg.message_type == "assistant_message":
        print(f"Sales Director: {msg.content}\n")

print("Analysis:")
print("- CEO updated executive_dashboard (private) and cross_dept_projects (directors)")
print("- Directors see strategic directive in cross_dept_projects")
print("- CEO doesn't need to message each director individually")
print("- Demonstrates top-down information cascade through shared blocks")
```

<Accordion title="Expected Output">
```
======================================================================
SCENARIO 8: CEO Decision Cascades Down
======================================================================

User → CEO: 'The board wants us to prioritize the enterprise segment...'

CEO: I've updated our strategic direction based on the board meeting:

**STRATEGIC PRIORITY UPDATE: Enterprise Segment Focus**

Updated in executive_dashboard:
- Strategic Priority #1: Enterprise market expansion (elevated priority)
- Target: 5 new enterprise customers by EOY (was 3)
- Investment: Increased focus on enterprise features

Updated in cross_dept_projects:
- Added strategic directive for all directors
- Each department should align their work with enterprise focus

**Department Implications:**
- Sales: Prioritize enterprise pipeline, target larger deals
- Engineering: Fast-track enterprise features, improve onboarding
- HR: Hire profiles with enterprise SaaS experience

I've coordinated this through cross_dept_projects so all directors are aligned.
They can now cascade this down to their teams and update their department priorities
accordingly.

Now check if Directors see the update...

User → Sales Director: 'Any strategic updates from leadership?'

Sales Director: Yes, I see a significant strategic update from the CEO in the
cross-department projects block:

**NEW STRATEGIC PRIORITY: Enterprise Segment Focus**

The CEO has communicated that we're doubling down on enterprise customers. This affects
my sales strategy immediately:

**Sales Implications:**
✓ Align perfectly with our current focus (Acme Corp win today)
✓ Prioritize enterprise pipeline ($2M+ deal sizes)
✓ Adjust messaging and positioning for larger organizations
✓ May need to shift resources from SMB deals to enterprise

**Action Items for My Team:**
1. Brief Rep 1 and Rep 2 on enterprise prioritization
2. Review pipeline and identify enterprise opportunities
3. Coordinate with Engineering on enterprise feature roadmap
4. Update sales_knowledge with enterprise-focused playbook

This is great strategic clarity from leadership. Our Enterprise Customer Onboarding
project with Engineering just became even more critical.

Analysis:
- CEO updated executive_dashboard (private) and cross_dept_projects (directors)
- Directors see strategic directive in cross_dept_projects
- CEO doesn't need to message each director individually
- Demonstrates top-down information cascade through shared blocks
```
</Accordion>

## Part 4: Block Access Matrix and Summary

```python
print("\n" + "=" * 70)
print("COMPREHENSIVE BLOCK ACCESS MATRIX")
print("=" * 70)

# Create comprehensive access report
blocks_to_check = [
    (company_mission, "company_mission"),
    (company_policies, "company_policies"),
    (sales_knowledge, "sales_knowledge"),
    (engineering_specs, "engineering_specs"),
    (hr_employee_data, "hr_employee_data"),
    (cross_dept_projects, "cross_dept_projects"),
    (executive_dashboard, "executive_dashboard"),
]

for block, name in blocks_to_check:
    print(f"\n{name}:")
    block_info = client.blocks.retrieve(block_id=block.id)
    print(f"  Agents with access: {len(block_info.agent_ids)}")
    for agent_id in block_info.agent_ids:
        agent_info = client.agents.retrieve(agent_id=agent_id)
        print(f"    - {agent_info.name}")

print("\n" + "=" * 70)
print("TUTORIAL 4 COMPLETE: Key Takeaways")
print("=" * 70)

print("""
✓ Created complete enterprise system with 10 agents across 4 organizational levels
✓ Implemented 4-tier block hierarchy (company → department → cross-dept → executive)
✓ Demonstrated department isolation (Sales/Eng/HR cannot see each other's data)
✓ Showed cross-department coordination via cross_dept_projects
✓ Demonstrated executive oversight via executive_dashboard
✓ Showed bidirectional information flow (top-down and bottom-up)
✓ Demonstrated HR confidentiality protection
✓ Showed asynchronous coordination eliminating explicit messaging

COMPLETE BLOCK ACCESS MATRIX:
==============================
                        company_ company_ sales_   engineering_ hr_      cross_   executive_
                        mission  policies knowledge specs       employee_ dept_    dashboard
                                                                data      projects
CEO                     R        R        -        -            -         R/W      R/W
Sales Director          R        R        R/W      -            -         R/W      -
Sales Rep 1             R        R        R/W      -            -         -        -
Sales Rep 2             R        R        R/W      -            -         -        -
Engineering Director    R        R        -        R/W          -         R/W      -
Engineer 1              R        R        -        R/W          -         -        -
Engineer 2              R        R        -        R/W          -         -        -
HR Director             R        R        -        -            R/W       R/W      -
HR Rep 1                R        R        -        -            R/W       -        -

R = Read-only | R/W = Read/Write | - = No Access

ARCHITECTURAL PATTERNS DEMONSTRATED:
====================================
1. **Company-wide blocks** (read-only): Universal access, consistent messaging
2. **Department isolation**: Each dept has private knowledge base
3. **Cross-department coordination**: Directors collaborate via shared block
4. **Executive aggregation**: CEO sees metrics without micromanaging
5. **Hierarchical information flow**: Top-down (strategy) and bottom-up (reporting)
6. **Confidentiality boundaries**: HR data strictly protected
7. **Asynchronous coordination**: No explicit messaging needed between departments

SCALING INSIGHTS:
=================
- 10 agents with 17 total blocks (7 shared + 10 private)
- Each agent has 3-5 block access on average
- Complex access patterns scale well
- Department boundaries maintain security
- Cross-functional work enabled without breaking isolation

REAL-WORLD USE CASES:
=====================
✓ Enterprise organizational structures
✓ Regulated industries (HIPAA, GDPR compliance via block isolation)
✓ Multi-team product development
✓ Customer support with specialized teams
✓ Educational institutions (departments, admin)
✓ Government agencies (clearance levels, department boundaries)

COMPARISON ACROSS ALL TUTORIALS:
=================================
Tutorial 1: Hierarchical access (3 tiers, read-only knowledge)
Tutorial 2: Uniform team access (task coordination, read/write)
Tutorial 3: Overlapping access (5 specialists, mixed blocks)
Tutorial 4: Organizational hierarchy (10 agents, 4-tier blocks, department isolation)

KEY TAKEAWAY:
=============
Shared memory blocks enable enterprise-scale multi-agent systems with:
- Appropriate information boundaries
- Asynchronous coordination
- Hierarchical oversight
- Department autonomy
- Cross-functional collaboration

All without explicit agent-to-agent messaging or centralized control!
""")
```

<Accordion title="Expected Output">
```
======================================================================
COMPREHENSIVE BLOCK ACCESS MATRIX
======================================================================

company_mission:
  Agents with access: 10
    - CEO
    - Sales_Director
    - Sales_Rep_1
    - Sales_Rep_2
    - Engineering_Director
    - Engineer_1
    - Engineer_2
    - HR_Director
    - HR_Rep_1

company_policies:
  Agents with access: 10
    - CEO
    - Sales_Director
    - Sales_Rep_1
    - Sales_Rep_2
    - Engineering_Director
    - Engineer_1
    - Engineer_2
    - HR_Director
    - HR_Rep_1

sales_knowledge:
  Agents with access: 3
    - Sales_Director
    - Sales_Rep_1
    - Sales_Rep_2

engineering_specs:
  Agents with access: 3
    - Engineering_Director
    - Engineer_1
    - Engineer_2

hr_employee_data:
  Agents with access: 2
    - HR_Director
    - HR_Rep_1

cross_dept_projects:
  Agents with access: 4
    - CEO
    - Sales_Director
    - Engineering_Director
    - HR_Director

executive_dashboard:
  Agents with access: 1
    - CEO

======================================================================
TUTORIAL 4 COMPLETE: Key Takeaways
======================================================================
... (full summary as shown above)
```
</Accordion>

## Key Takeaways

<CardGroup cols={2}>
<Card title="4-Tier Hierarchy" icon="layer-group">
Company → Department → Cross-Dept → Executive blocks create organizational structure
</Card>

<Card title="Department Isolation" icon="shield-halved">
Each department maintains private knowledge bases invisible to other departments
</Card>

<Card title="Asynchronous Coordination" icon="arrows-spin">
Directors coordinate via cross_dept_projects without explicit messaging
</Card>

<Card title="Executive Oversight" icon="eye">
CEO sees aggregated metrics and coordination without micromanaging
</Card>
</CardGroup>

### Architectural Patterns

| Pattern | Description | Blocks |
|---|---|---|
| **Universal Access** | All agents see | `company_mission`, `company_policies` |
| **Department Silos** | Department-only access | `sales_knowledge`, `engineering_specs`, `hr_employee_data` |
| **Director Coordination** | Directors + CEO | `cross_dept_projects` |
| **Executive Private** | CEO only | `executive_dashboard` |

### Comparison Across All 4 Tutorials

| Aspect | Tutorial 1 | Tutorial 2 | Tutorial 3 | Tutorial 4 |
|---|---|---|---|---|
| **Agents** | 4 | 4 | 5 | 10 |
| **Shared Blocks** | 3 | 3 | 5 | 7 |
| **Access Pattern** | Hierarchical tiers | Uniform team | Overlapping | Organizational hierarchy |
| **Complexity** | Low | Medium | High | Very High |
| **Read/Write Mix** | Read-only | Read/Write | Mixed | Mixed |
| **Use Case** | Support tiers | Task coordination | Personal assistant | Enterprise org |

## Series Conclusion

You've now completed all 4 shared memory block tutorials! Here
's what you've learned:

### Tutorial Progression

**Part 1: Read-Only Knowledge** → Foundation
- Hierarchical access control
- Read-only blocks for policies
- Information consistency

**Part 2: Task Coordination** → Collaboration
- Read/write shared blocks
- Worker coordination patterns
- Knowledge sharing

**Part 3: User Assistant** → Specialization
- Overlapping access patterns
- Privacy boundaries
- Zero-handoff workflows

**Part 4: Enterprise System** → Scale
- Organizational hierarchies
- Department isolation
- Cross-functional coordination
- Executive oversight

### Universal Principles

Across all tutorials, these principles hold:

1. **Appropriate Access**: Agents only access blocks they need (least privilege)
2. **Real-Time Sync**: Updates to shared blocks immediately visible to all agents with access
3. **No Explicit Coordination**: Shared blocks eliminate need for agent-to-agent messaging
4. **Privacy Boundaries**: Sensitive data restricted to appropriate agents
5. **Scalability**: Patterns work from 2 agents to 10+ agents

## Next Steps

<CardGroup cols={2}>
<Card title="API Reference" icon="code" href="/api-reference/agents/core-memory">
Complete API documentation for blocks and memory management
</Card>

<Card title="Multi-Agent Guide" icon="users" href="/guides/agents/multi-agent">
High-level overview of multi-agent system architectures
</Card>

<Card title="Shared Memory Guide" icon="database" href="/guides/agents/shared-memory">
Comprehensive guide to shared memory blocks (coming soon based on these tutorials!)
</Card>

<Card title="Discord Community" icon="discord" href="https://discord.gg/letta">
Get help and share your multi-agent implementations
</Card>
</CardGroup>