Answering Shipping Customers on WhatsApp With an AI Agent: A Playbook

This is a playbook, not a client story. Where it uses numbers, they are an illustration to show the arithmetic: put your own in.

Small shipping businesses, including the many that serve diaspora communities, often run customer service on WhatsApp. Before holidays the same questions arrive in waves: where is my shipment, when will it arrive, what does it cost to send to this city, what can I send. A small team answering all of them has no time left for the problems that need a person.

Check whether an agent will help

For one week, tag every incoming question by type. If most of them fall into a short list of repeat questions, and the answers live in data you already hold (tracking, routes, prices, pickup hours, item rules), an agent can take the repeats. If most questions are one-off problems, it will not.

What to connect

• WhatsApp Business Platform and your website chat, so customers stay where they are.
• Your tracking data, read-only, so status answers are real and not guessed.
• A written price list, route list and item rules the agent quotes from, never invents.

When it hands over

Claims, damage, customs problems, complaints and anything involving money back go to a person, with the conversation summarised so the customer does not repeat themselves. Make the hand-over easy to trigger: a customer typing "agent" or sounding upset should reach a human.

Measure it

Track first-response time, the share of conversations resolved without a person, repeat contacts about the same shipment, and complaints about wrong answers. Illustration: if 200 messages a day arrive before a holiday and an agent resolves half, your team handles 100 instead of 200. Your share will depend on how many questions are truly repeats.

Practical takeaway: Do the one-week tagging exercise first. It tells you whether to build an agent, and it produces the question list the agent is built from.