Answer the question before trying to sell
Someone asking about a size, measurement or shipping method needs a specific answer. A chatbot for an online shop in Spain can help locate that information without searching the entire catalogue. Its first task should be to understand the question and find a valid response.
Review existing team enquiries and choose a small group to start. Product features and published conditions can provide a useful foundation. If an answer depends on an exception or commercial commitment, identify who needs to review it before replying.
Maintain a reviewed information source
The bot needs to know which content it may use and who updates it. An outdated product record may describe a price or delivery estimate the shop has changed. Assign owners for the catalogue, euro prices and delivery conditions, and decide how changes reach the answers.
If choosing an AI solution, limit answers to authorised information and provide a route for unknown questions. Avoid filling a gap with a plausible policy that does not exist in the business. Accuracy matters more than replying to every question.
Separate general information from order details
Explaining supported destinations differs from locating a specific parcel. The first can use public conditions. The second needs order system access and an appropriate check before exposing information belonging to a person.
Do not request sensitive documents routinely or disclose information merely because someone knows an order number. Design checks with the security and privacy owners. Without a reliable integration, the bot can point to the existing tracking channel or hand over the enquiry.
Escalate incidents with a clear explanation
A complaint, damaged product or special request should not become trapped in a menu. Let people request a person and pass on the relevant context. The bot can recognise the reason, explain necessary information and register the case for review.
Avoid promising refunds, replacements or exceptions the system cannot authorise. An automatic response can explain the shop's process; resolution must follow agreed procedures. Give opening hours and a response expectation the team can fulfil.
Make the conversation straightforward
Use short messages, understandable choices and a way to go back. When customers use their own words, confirm the system's interpretation before taking an action. Do not make people restart every time an answer falls outside the expected journey.
Identify automated service and explain how to request human help. WhatsApp communications must follow channel permissions and rules. Coordinate requirements with the business owners rather than assuming every conversation allows subsequent promotional messages.
Evaluate resolution and customer effort
Measure whether questions were resolved and whether people had to repeat their information to an agent. Review abandonment and nonsensical answers. Counting messages alone cannot establish whether the experience was helpful.
Expand once answers, sources and handovers work. Review content whenever the catalogue or logistics change. A chatbot adds value by making information easier to find and giving the team better context for cases that require human judgement.
