The most common uses of chatbots in ecommerce are to answer pre-purchase questions. These are questions that a customer might ask before they make a purchase. So questions about the products, about shipping policies, or returns. It includes basically anything that would make a customer more confident about a purchase.
Incidentally – these aren’t the most common queries that customers tend to ask, for most retailers. Those would be questions about order status, or WISMO queries (Where is My Order).
There are a number of different ways you can look to use customer service AI to deal with pre-purchase queries. Let’s take a look at how different retailers approach to answering one particular question: “How much is shipping to the UK”?
Collect the answer to respond to later
One approach we’ve seen is that the chatbot is unable to deal with the question, and so acts a bit like a voicemail machine, taking down the question and passing it to an agent to pick up later.
Obviously this approach is not ideal. It could be hours, or even days until the question is answered, by which time the customer may have completely moved on and forgotten all about what they were even looking at.
The one positive is that the AI hasn’t tried to answer the question and got it wildly wrong. Again this is something we’ve seen with some bots getting confused and giving an answer that makes no sense.
Immediately hand over to an agent
This approach is a step further than taking a message and involves a live chat handover to an agent. If the handover is successful, this is obviously preferable to taking a message because the customer gets an answer within a few messages, while staying on the site.
Once again, it’s a positive that the AI didn’t try and answer a question it wasn’t able to. What was less good about this example was it took the agent a long time to respond, and when they did respond it was a vague and unhelpful message.
“We offer free shipping if you reach a certain threshold.” - what is the threshold?
“Shipping will depend on the method selected.” - that’s all well and good, but perhaps you could explain it to the customer?
Send the customer to a help article
This is one of the most common ways of dealing with pre-purchase questions is to send customers to a help article or FAQ page that should answer their question, as in the example below.
To be effective, you still need to be able to fully understand what they are saying in order to find the right page to send them to the right place. In this example the bot has nearly got it right, but has taken the customer to the wrong page, one that has a similar topic but is different.
If it had taken the customer to the right place, it would still hae been asking the customer to read through a potentially long page of text to get to the answer they were looking for. That’s not a fantastic customer experience, but if it works it does do the job.
Send a templated response based on keywords
Rather than sending customers to a page, a better experience is to summarise the answer to their question in a templated response, as in the example below.
If a customer has a specific question then it’s clearly a better experience to answer that question specifically, rather than sending them to a page where the answer might be.
Plus of course, you can keep customers on page and keep answering their questions.
In the example, the chatbot didn’t quite nail the answer, but gave the best response so far. It gave some relevant information about shipping, but didn’t give a figure for shipping in the UK.
Respond using generative AI
Generative AI can go a step beyond the templated response and give a more personalised response and also get to the heart of the question.
In the example below, the AI is able to look up the information on the relevant page, digest it and then rephrase it for the customer. In this case it gives an accurate answer.
This does require using an AI model that understands the intent in the message and checks whether it knows the answer. There are a number of steps to be taken to ensure good generative AI for customer service, which we’ve covered here.
The obvious plus points are that this actually answers the question, and it answers it within the conversation, meaning that the shopping journey is not interrupted. With this information, a barrier to purchasing has been removed and customers can move on to the next thing. They’re also not having to do lots of work looking for the answer.
Talk to us to find out how you can use a chatbot to answer pre-purchase questions.
