How to Train Your Chatbot

AI and humans make a great team; we know this partnership is the future of work.

Like human employees, digital employees, or AI chatbots, need structured ongoing training, ensuring they continue to learn and improve their ability to answer nuanced questions. This is vital when a chatbot is rolled out to make sure it learns industry-specific terms and processes.

While responsibility for training chatbots might initially sit within an IT or customer services team, we believe this emerging discipline is of interest to, and can itself learn from, the learning and development and CRM professionals.

To maximize the benefits of AI technology, it's important that a chatbot is integrated with other systems, such as the CRM system, e-commerce platform, and customer service tools. This will ensure that the chatbot has access to the data it needs to provide the best possible marketing and customer experience.

A digital employee needs ongoing training and practice, just like a regular employee. This is a pretty straightforward process that involves the following steps:

  • Assign a champion. Ideally this is someone recruited for a dedicated support role to train the digital employee.
  • After the initial setup, the champion continues to train the chatbot on five to 10 conversations each week, ensuring that they continually learn and respond correctly to the questions customers are asking.
  • When something new occurs that might trigger an influx of inquiries, such as a new product launch, the champion quickly brings the digital employee up to speed on the new answers it will need.
  • The chatbot should quickly learn what people want and will give suggestions for which info can be added next. On a good chatbot system it's really easy!

How the Chatbot Can Help Train Your Humans

While training the chatbot is crucial, the flipside of that opportunity is equally interesting: how AI can be used to help train and onboard new staff. Following are a few ideas:

  • Employees can access company internal documentation and processes by asking the digital employee for it in the chat.
  • The chatbot can help employees understand product information and stock levels without the need for extra systems.
  • New employees can be onboarded with chatbot training sessions; it's good for both parties as they learn together.

How Companies Succeed with AI Chatbots

Ambit's digital employees are used by some of New Zealand's and Australia's largest companies, including Laybuy, Tower Insurance, and Glassons. These organizations share many of the same challenges: providing a consistent level of customer service 24/7, even during periods of high demand, whether that's due to a sudden weather event or an online sale.

Laybuy, New Zealand's leading buy-now-pay-later (BNPL) provider, expanded into Australia and the United Kingdom just as the popularity of BNPL was about to explode due to the pandemic. The company quickly found itself with close to a million customers and a growing demand on its customer support team.

As it headed into the busy 2021 Christmas period, the situation was unsustainable, and it hired its first digital employee, Hugo (short for Helping Users Get Organized). Within three weeks, Hugo was answering 7 percent of Laybuy's customer inquiries, equating to around 1,400 queries a day.

Today, Hugo is a key member in Laybuy's customer support team, answering more than 70 percent of inquiries, providing instantaneous answers to common everyday questions, such as "How do I reset my password?" or "How do I return an item?"

By managing these high-volume but simple inquiries, Hugo allows the rest of the support team to focus on customers with complex questions that need dedicated assistance. This is important because, in a competitive retail environment, reputation is everything. If you don't look after your customers quickly you risk them taking their business elsewhere.;

Hugo puts Laybuy in a better position to meet and exceed customer expectations, delivering on customer service every time they make contact. Response times have been reduced by 75 hours, and customer satisfaction increased by 30 percent.

But Hugo also helped deliver significant savings for the business. Prior to using Hugo, the only way Laybuy could meet the growing demand on its support team was to employ more staff, who are not only increasingly scarce but come at a significant cost. AI chatbots, such as Hugo, by comparison, have a cost to serve of only 47 cents per inquiry and can help reduce customer service costs by nearly a third.

The success of any AI chatbot can also be seen in the deflection rate—the percentage of emails deflected to the customer support team because the chatbot couldn't provide an answer. During the first year of Hugo learning his job, the monthly number of support tickets deflected to the team fell by 61 percent.

Another real test was the yearly Laybuy Mania sales event. During the 24-hour sale, the company processed more than 21,500 transactions, but received fewer customer support tickets than at the same event last year.

A key to Hugo's success was the commitment to training him internally, using the champion system. As Laybuy continues to grow, Hugo can grow with it.


Tim Warren is co-founder and CEO of Ambit.