All case studies

E-commerce retailer, United States

A US E-Commerce Retailer: Handling High-Volume WhatsApp Support Without Losing the Personal Touch

A US e-commerce retailer replaced manual WhatsApp support with an AI agent trained on company data, with escalation to human agents. Response time fell by 90 per cent.

Timeline: Six-week deployment roadmap

• 90% Reduction in response time
• 85% Improvement in response accuracy and relevance
• 95% Routine inquiries automated
• 400% Increase in simultaneous conversation capacity
• $300K+ Annual savings in customer service costs

The problem

The support team was overwhelmed by high-volume WhatsApp inquiries. Human agents handled routine questions by hand, so replies were slow and inconsistent. Traditional chatbots gave generic answers without company-specific knowledge, which drove up escalations, wait times and costs, and the existing setup could not meet demand around the clock.

• Support staff spent too long on repetitive, routine questions
• Generic chatbot replies lacked company-specific knowledge
• The support setup could not meet 24/7 demand or keep a consistent brand voice
• Lack of context drove high escalations and longer waits

What we built

We built a WhatsApp support system on the WhatsApp Business API. Each message triggers a Make.com workflow that passes it to a ChatGPT agent hosted on Amazon Lambda and trained on the retailer's own company data. The agent uses the customer's conversation history to write a personalised reply, which Make.com sends back. Every conversation then goes through an escalation check: if a person is needed, it is flagged and the human support team is notified, otherwise the automated conversation continues and the history is updated.

Tools: WhatsApp Business API, Make.com, Amazon Lambda, ChatGPT, Conversation history database

How we ran it

Every engagement follows The Claro Build Framework.

1. Assess: We reviewed how WhatsApp inquiries were handled and found routine questions consuming agent time and generic chatbot replies driving escalations.
2. Design: We designed an AI agent grounded in company data, context handling from conversation history and clear escalation rules for handing over to a person.
3. Build: We trained the AI on company data and integrated the WhatsApp API, set up Lambda, routing and context management, then tested the escalation rules before full launch.
4. Sustain: Handover followed The Adoption Standard: a structured 45-minute handover call and documentation written as the system was built.

The result

Customer service response time fell by 90 per cent and response accuracy and relevance improved by 85 per cent. Routine inquiries are now 95 per cent automated, capacity for simultaneous conversations rose by 400 per cent with round-the-clock support, and the retailer reports more than $300,000 in annual savings.

Client name withheld for confidentiality.