AI Email Generation

The M365 Knowledge Worker workload supports AI-powered email content generation using the LLM abstraction layer from agent-haymaker.

Overview

When enable_ai_generation is set to true, the workload uses an LLM provider to generate realistic, department-appropriate email content. When the LLM is unavailable, it falls back to template-based content.

Enabling AI Email Generation

Via CLI

haymaker deploy m365-knowledge-worker \
  --config workers=25 \
  --config department=sales \
  --config enable_ai_generation=true \
  --config email_directive="Write about Q4 targets"

Via Config File

# ai-email-deployment.yaml
workload_name: m365-knowledge-worker
workers: 10
department: sales
duration_hours: 4
enable_ai_generation: true
email_directive: "Write emails about Q4 sales targets and client relationships"

Configuration

Field	Type	Default	Description
enable_ai_generation	bool	false	Enable LLM-powered content
email_directive	string	null	Custom directive for email topics

How It Works

1. Deploy phase: Workload checks enable_ai_generation and creates LLM client
2. Activity phase: Each email activity uses EmailGenerator:
   • If LLM available: generates content via LLM with department-aware prompts
   • If LLM fails: falls back to template-based content
   • Logs generated subject line for observability

Department-Aware Prompts

Email content varies by department:

Department	Typical Topics
Engineering	Code reviews, sprint updates, technical decisions
Sales	Client meetings, deal progress, quarterly targets
HR	Policy updates, onboarding, team events
Finance	Budget reviews, expense reports, financial planning
Operations	Process improvement, vendor coordination, logistics
Executive	Strategic initiatives, board preparation, organizational updates

Custom Directives

The email_directive parameter lets you customize the email generation:

# Limerick-style emails
email_directive: "Write all emails as limericks"

# Specific business context
email_directive: "Emails should reference the upcoming product launch"

LLM Provider Setup

AI email generation requires an LLM provider. Set environment variables before deploying:

# Anthropic
export LLM_PROVIDER=anthropic
export ANTHROPIC_API_KEY=sk-ant-...

# Or Azure OpenAI
export LLM_PROVIDER=azure_openai
export AZURE_OPENAI_ENDPOINT=https://myresource.openai.azure.com
export AZURE_OPENAI_DEPLOYMENT=gpt-4

Install AI dependencies:

pip install haymaker-m365-workloads[ai]

Fallback Behavior

If the LLM is unavailable (not configured, API error, rate limited), the system falls back to rotating department-specific email templates. This ensures telemetry generation continues even without AI.

Architecture

haymaker_m365_workloads/
├── content/
│   ├── email_generator.py   EmailGenerator (LLM + fallback)
│   └── prompts.py           Department-aware prompt templates
├── workload.py              LLM client creation on deploy
└── operations/
    └── orchestrator.py      Uses EmailGenerator for email activities