Product Knowledge AI Chatbots

I’m going to continue discussing the Product Knowledge AI chatbot.
The AI chatbot needs to be trained on your products. “Training” is as simple as uploading a document describing the product, its uses and its benefits. This comes from your website and product brochures. It’s a 1-minute process to upload.
Multiple documents or website pages can be uploaded. At any time additional pages can be uploaded, or removed. It does not need to even follow a specific format. The AI behind it (the LLM) can match any question to the right answer.

At this point the AI agent is still not perfect. Customers may ask questions that are not on the website or in the product information fact sheet. Examples could include do you have stock or how long will it take to deliver the product to me. Your customer may also have some specific requirements that are not in the fact sheet or on the website. That is generally why many companies prefer to have a human answering the query than an AI product knowledge chat bot. They are of course wrong. Many sales staff also do not have the required information and would have to go back to the technical manager or the sales manager for that information.

In the case of an AI chatbot if the chatbot does not have the required information the customer can choose to get the AI chatbot to send a message to the company asking for more information at which point senior management would reply with the required information. The benefit of an AI chatbot is management can see and monitor every question asked and every answer given and judge whether the answer is correct. If the answer is not perfect a person in the company can very quickly update the answer using our back-end software. For example if a customer says how long will it take to deliver and the chatbot does not have the answer an employee could update the answer with “it generally takes three days to courier down to Cape Town”. At that point and from then on the AI chatbot will have this information for the next time a customer asks a similar question. This is how machine learning works. In our experience it would take no more than a couple of weeks for the AI chatbot to have 90% of all the answers, ie a lot more answers than a call centre employee or a junior sales person would have.

A point to note is that once the AI chatbot has been trained it does not need training on the same information again unlike a new employee would require training. If a new product is launched the usual process would be to get all the product managers and support agents into training rooms and train them on the new product. In the case of AI, one document would train the system no matter how many questions and answers it handles.