Case study
Zawadi AI Support Agent
AI customer support automation · Production
A customer support agent built for a live ecommerce business, connecting customer conversations with order lookup, human escalation and support logging.
- Status
- Live in production
- Contribution
AI agent design
n8n workflow development
Claude integration
Shopify integration
Support automation
01 — The problem
Customer questions should not end in another manual inbox.
The store already had a functioning ecommerce experience. The missing layer was support that could respond inside the shopping journey and continue into a real action when the question required store data or a human decision.
The goal was therefore broader than a chatbot: create a support workflow that could answer, retrieve, escalate and leave a usable record behind.
02 — The intervention
Support moved onto the storefront.
Zawadi was added as a branded customer-facing support layer on Kenyan Store Deutschland. The interface gives shoppers a direct place to ask for help without leaving the store.
03 — Customer experience
A simple front end, backed by a real workflow.
The shopper sees a straightforward chat. Behind that interface, the agent has memory and tools, so the conversation can move beyond a generated reply when the request requires an order lookup or owner intervention.
Portfolio conversations use sanitized demonstration data. No private customer or order information is shown.
04 — How Zawadi works
From customer question to business action.
AI interprets the request. The workflow controls the action.
- 01
Customer message
A customer asks Zawadi a support or order-related question.
- 02
AI interpretation
Claude interprets the customer's request and determines what information or action is needed.
- 03
n8n orchestration
n8n manages the workflow and gives the AI agent access to the appropriate connected tools.
- 04
Shopify order lookup
When order information is required, the workflow can retrieve relevant order data through the Shopify integration.
- 05
Customer response
Zawadi uses the available context and retrieved information to respond to the customer.
Human escalation
Requests requiring owner attention can be escalated through Gmail.
Interaction logging
Support interactions can be logged to Google Sheets.
The AIClaude interprets the customer's request and determines what information or action is needed.
The workflown8n and the connected business systems handle workflow execution and any external action.
05 — Real implementation
The automation behind the diagram.
The n8n workflow connects the chat trigger to the support agent, Claude, conversation memory and the actions the agent can take.
06 — Agent actions
The answer is only one part of the job.
- 01
Answer
Handles customer questions through a branded storefront chat experience.
- 02
Remember
Conversation memory keeps the interaction coherent across the session.
- 03
Look up
The agent can call the Shopify order-lookup tool when an order question needs real store data.
- 04
Escalate
Questions that need a person can be sent to the owner by email instead of being guessed.
- 05
Log
Support conversations are recorded in Google Sheets for visibility and follow-up.
07 — Why this project matters
AI connected to an operating business.
Zawadi demonstrates an end-to-end AI workflow rather than a standalone chat demo: customer interface, model reasoning, memory, ecommerce data access, human escalation, logging and a live production deployment.
08 — Build status
Live in production.