Valérian de Thézan de Gaussan · Data Engineering for process-heavy organizations

Stop using AI to create Customer Service chatbots, you're only upsetting your customers.

What you should do instead 👇


What’s more frustrating that a robot that cannot answer your question, is looping constantly over the same FAQs links, and is gatekeeping you from talking to an actual human, that could probably fix your problem in 20 seconds?

The typical symptom for this is that customer support is treated as a cost center, when it is actually a value-making department (it improves LTV, thus augmenting your CAC to LTV ratio -> more value for the same cost).

I keep my home insurance contract because the customer service of my insurance company is so great. I never looked at any competitors prices because of that. I even moved my car insurance contract to them even if it didn’t make sense cost-wise. I just know that I’ll have a great service with them.

Instead of using AI as a first-line barrier, it can be repositioned as a support system that enhances the capabilities of human agents. Here’s how AI can positively impact customer service productivity without replacing the human touch:

  1. Use an AI Agent on the customer service representative side, not the client side. AI can quickly go through data to provide CS representatives with relevant information, so they help address customer needs more effectively on complex problems. This will also remove issues of AI Chatbots going wild in front of customers, with all the legal issues it brings.

  2. By gathering and analysing data on past interactions, you will slowly build a knowledge base of FAQs. Gen AI can help you build it. Display that on your website, while making sure the UX is good enough. Monitor the amount of calls made for problems that are already registered in the FAQs, as this is a good indicator that your FAQs UX is not good enough.

  3. If you really want to use a chatbot, make sure it is only for initial screening, and redirect to human agent as soon as it figures it is not capable of answering.

  4. Setting follow-ups after resolving a customer’s issue. AI can craft custom follow-ups to make sure their exact problem is fixed. Good opportunity to do some upsells, and can nearly be fully automated.

AI should help, not hinder. Train representatives to use AI effectively. Always allow customers to reach a human if needed.

So, instead of using a chatbot to reduce the number of calls to your customer services, thus justifying decisions to cut costs, use AI to support human agents, not replace them.

Let’s get rid the world of the canned answers, the maze of chatbots who never wants to give you access to a human, while also reducing the risk of bringing Gen AI to the front (increased surface attack, legal issues etc…).

👉 Contact me if you want to talk about how you could save time and effort, while improving the quality of your output, using data and AI.

PS: The picture below is from Upwork support forum. The thread starts about the customer complaining about the chatbot going in loops. The responses from CS are copypasta with no care about the user’s issue. After a lot of back-and-forth, it took the support 2 weeks to fix an issue that could have been handled in 15 minutes. Don’t be like that.

stop-using-ai-chatbots-upwork.png