Week 2 connects your workflows to the platforms Kenya actually runs on. M-Pesa, WhatsApp Business, and Africa's Talking SMS are at the centre — giving your future AI agents the channels they need to operate in the real Kenyan market.

• Safaricom Daraja API: STK Push, C2B, B2C, B2B — which to use when
• Setting up sandbox credentials and running your first test STK Push payment
• Receiving M-Pesa IPN (Instant Payment Notifications) via n8n webhook
• Parsing IPN data: transaction code, amount, sender name, and phone number
• Triggering downstream actions immediately on payment confirmation
• WhatsApp Cloud API (Meta) vs Twilio vs 360Dialog — options explained
• Sending templated and free-form WhatsApp messages via n8n
• Building a keyword-based WhatsApp auto-responder without AI
• Africa's Talking: bulk SMS, USSD menus, and voice for rural Kenya coverage
• Airtable as a no-code CRM — create, read, update, delete records from n8n
• Lead capture pipeline: website form → Airtable CRM → auto email confirmation
• Advanced error handling: retry logic, fallback notification paths, dead-letter logging

This is where your workflows get intelligence. Week 3 introduces Claude AI and GPT-4 into your n8n automations — moving from simple rule-based bots to AI agents that can understand, classify, generate, and respond to real human messages in both English and Swahili.

• What is a Large Language Model? Plain-English explanation using Kenyan analogies
• GPT-4 vs Claude AI — capabilities, pricing, and when to use each
• Setting up OpenAI and Claude API credentials as n8n nodes
• The n8n AI Agent node: architecture, inputs, outputs, and system prompts
• Prompt engineering: system prompts, few-shot examples, and structured JSON output
• Multi-language prompts: English and Kiswahili hybrid responses for Kenyan audiences
• AI content creation: social media captions, product descriptions, and ad copy automation
• Intelligent classification: routing emails, tickets, and WhatsApp messages by topic and urgency
• Sentiment analysis: detecting positive, neutral, frustrated, or escalation-risk customers
• Information extraction: pulling names, amounts, dates, and entities from unstructured text
• Conversational agents with persistent memory — the n8n AI Agent node with memory enabled
• Session management: separate conversation threads per user, with 30-minute inactivity clearing

Week 4 takes your agent from a working prototype to a production-ready system deployed on WhatsApp. You will add a business knowledge base (RAG), give your agent the ability to take real actions, deploy it securely, and learn how to package and sell what you have built to Kenyan clients.

• RAG (Retrieval-Augmented Generation): teaching AI your business documents and policies
• Creating a vector knowledge base from PDFs, Word docs, and Google Docs
• Semantic search: customers ask questions, AI answers from your actual documents
• Tool-calling agents: AI that queries Google Sheets, books Calendar appointments, and sends WhatsApp messages mid-conversation
• Safety guardrails: preventing your agent from performing unauthorised actions
• Setting up a verified WhatsApp Business number with Meta Cloud API
• Connecting WhatsApp webhook to your n8n production workflow end-to-end
• Handling rich media: receiving images, documents, and voice notes in WhatsApp
• Human escalation: seamless AI-to-human handoff with full conversation context transfer
• Production deployment: n8n Cloud vs self-hosted on DigitalOcean — Kenya pricing compared
• Error monitoring: WhatsApp alerts when any workflow fails, with automated 3-retry logic
• Selling AI chat agent services: ODSAI pricing guide — WhatsApp AI Agent KES 30,000–80,000 per setup, KES 10,000–30,000/month retainer