The company: A small ecommerce business focused on CBD products.The industry: The CBD world is rapidly growing as cannabis/hemp products reach a mainstream audience. Due to a history of misinformation and confusion with THC & CBD products, there is a huge amount of customer education required in this area. This can be a roadblock for small companies that don’t have the resources to run a large support team to answer questions 24/7.The problem: The CBD startup found that they had been spending too much on driving traffic to their site without getting the results they were expecting. If customers don’t feel confident in what they are buying and who they are buying from, they are unlikely to stick around and read long pages of FAQs or product descriptions.The solution: A 24/7 product and service expert who can answer questions, provide recommendations and offer support. Sounds expensive and time consuming, right? Not really, actually. We were able to build an AI chatbot that learns from old chatlogs, FAQs, and support emails to become a product expert. Using a technique called Natural Language Processing, the bot is able to understand the intent of the message or questions.For example, it can pick simple things up like “What is hemp?” or “what does Sativa mean” and associates it with the Intent: Understand Cannabis Terminology.It can also take on more complex things depending on how you format and train your model. For example, “Can you show me some tinctures that can help fight my anxiety?”. This is a cool example because of how it can be broken down.The model parses the question and first identifies the overall intent of the question based on other user examples — in this case associating it with Intent: Get Product Recommendation.Then it parses for specific words (called entities) to associate to predefined categories. In this example, it is associating “tinctures” with the Entity: Consumption Method (versus edibles, vape, topical), and “anxiety” with the Entity: Ailment (versus things like sleep, pain, etc.). By doing this, we are able to tie this back to the company’s product feed and categorize their SKUs for the bot to be able to provide relevant product recommendations.Here is a graph showing the most common questions the chatbot got about CBDThe result: For this CBD startup we were able to automate 47% of their customer service inquiries. The instant response and 24/7 support has also increased conversions by 35%. The amazing thing about this is that it hasn’t resulted in an increased workload for their employees. As customers ask more and more questions, the bot gets smarter and smarter. We expect to be able to answer upwards of 80% of questions in certain applications. This also allows the bot to become an expert on any subject, not just CBD. We currently have a few projects in progress with apartment buildings, storage units, and travel agencies. If the bot doesn’t know the answer to a question it pings a real human and they connect in real time. During off hours it takes down the user’s info so sales or service can get in touch.Feel free to leave any questions you have about our chatbot or natural language processing in general!