“AI Chatbots don’t work” . Yes, that’s true, We hear that all day and we often say it ourselves.
On the other hand AI Chatbots do work very, very well. The differentiator is the implementation and management.
How Do we Make it Work?
It needs a proper plan and regular management. Just as you would never leave your call centre agents to their own devices after employing them, you cannot ignore your AI Chatbots. The difference is it’s a lot easier to train and then re-train one chatbot than ten call centre agents.
Before we go further, let us summarise the benefits of an AI agent:
- It operates 24/7. Imagine giving your customers support at 3am because that is when they need it.
- It scales. It can do the job of 1 or 10 or 100 or 1000 call centre agents. Think how long it takes to employ and train 1 new call centre agent.
- It learns. When it gets a support call wrong, it can learn from the updated answer that the supervisor gives.
- It can recognise (understand) and respond in multiple languages.
- It is fast. On average it gives an answer within 10 seconds. It can have a full conversation with a customer, fully solve the problem before a call centre agent has answered the telephone.
- It is always consistent, it is always polite, it is always professional. While it can come across as empathetic, it has no emotion to get upset, irritated or annoyed with customers no matter how they may behave.
Our Approach
Initial briefing:
We need to know how the business works
Understand the support issue:
There are typically various channels of support in a business:
– Email
– Telephone
– Forms on a website
– Walk-in customers
– Instant chat on a web site (non AI)
– WhatsApp/SMS
Understand the scope of support: How many support issues are generated and handled by channel per day, per week, per annum
Understand the type of support given: We will ask for a list of every support query and every support response. This will form the knowledgebase that the AI Chatbot will use.
Design the AI Chatbot: This refers to the overall look and feel of the chatbot which will be installed on a web site that customers will have access to. In addition an AI WhatsApp will be created and the mobile number given to customers
Update the knowledgebase: The better the knowledgebase the better the AI Chatbot. The knowledgebase will contain details of the company and its products. as mentioned above the knowledgebase will contain details of previous queries and responses. It may contain user manuals
Testing: Testing will take place on an on-going basis – just like call centre agents are assessed regularly. All responses will be monitors and additional knowledge give to the AI Chatbot when it is required
Going live: As soon as the AI Chatbot is live all responses will be carefully monitored. While it is impossible to train the AI Chatbot for every eventuality it should be able to handle the vast majority of generic enquiries.
Regular monitoring and statistics: This is an on-going task. Over time the number of customers asking questions will increase as customers begin trusting the AI Chatbot. Areas where the AI Chatbot cannot answer will be investigated to provide a solution.