24/7 customer support without hiring more staff
In online stores with 2,000+ monthly orders, hundreds of repetitive queries arrive outside business hours. An AI chatbot trained on the full business documentation resolves 80% of queries instantly and escalates the rest with complete context.
The context
A natural cosmetics store handles over 2,000 orders per month. The support team — 3 people — works Monday to Friday from 9 AM to 6 PM, but 40% of queries arrive outside those hours: evenings, weekends, and holidays.
The most frequent queries are always the same: order status, return policy, product ingredients, shipping times, and stock availability. The team estimates that 70-80% of questions have a standard answer they repeat dozens of times daily.
The challenge
Customers who write outside business hours don't get a response until the next day — sometimes up to 16 hours later. In e-commerce, that's an eternity. Many abandon their purchase or leave negative reviews mentioning the lack of support.
Hiring more staff to cover nights and weekends isn't economically viable. And chatbot solutions based on decision trees frustrate customers because they can't understand naturally phrased questions and end up in loops with no useful answer.
The solution
The chatbot is powered by the OpenAI API, integrated into the website and WhatsApp Business. It's trained on all business documentation: product catalog with ingredients and descriptions, shipping and return policies, FAQs, and usage guides for each product.
The chatbot understands natural language and can answer complex questions like "Is this serum suitable for sensitive skin and how long does it take to ship to Barcelona?" It also connects to the shipping platform's API to provide real-time information about specific order statuses.
When the bot detects it can't resolve a query with confidence — complaints, payment issues, special cases — it automatically escalates to a human agent. The agent receives a conversation summary with full context, so they don't have to ask the customer to repeat anything. Typical implementation takes 5 days.
Results
80% of queries are resolved without human intervention, including nights and weekends. Average response time goes from hours to seconds.
The support team goes from answering repetitive questions to handling only cases that truly need human attention. Job satisfaction improves notably because they solve interesting problems instead of copying and pasting the same answers.
Support-related negative reviews drop by 60% in the first quarter. Customers write at 2 AM and get a useful answer in 3 seconds — something impossible without automation.
Lessons learned
- Chatbot quality depends directly on the quality of the documentation it's trained on. Investing time in organizing and completing FAQs before launch multiplies its effectiveness.
- Escalation with context is as important as automatic resolution. An agent who receives the customer with the full history resolves the case in half the time.
- WhatsApp turns out to be the most-used channel — 65% of queries come through it. Being where the customer already is beats trying to get the customer to come to your website.