Chatbots have failed at customer service – where to next? 

by Ryan Falkenberg. Chatbots were coming to save the day in customer service. They promised automation, a better experience, and a way for businesses to learn.

by Ryan Falkenberg. Chatbots were coming to save the day in customer service. They promised automation, a better experience, and a way for businesses to learn from their bots’ engagements with their customers and improve over time. And that’s all good and well. The thing is, chatbots are all about natural language and unstructured conversations. And most businesses live in a structured world, with defined processes and procedures, and right or wrong answers to customer queries.

If your chatbot is learning how, for example, to predict your next purchase based on past behaviour, and it gets it wrong, it’s no biggie. If the bot is guiding financial services customers on products or investments, on the other hand, and it doesn’t follow the correct process or gives the wrong advice, we have a massive problem. Bots engaging with customers on service queries need to follow set procedures – specific questions need to be asked, rules need to be applied, in a specific order, so that the outcome can then be passed to other digital workers (or humans) to execute.

It’s important if it gets something wrong. If it is working off past customer behaviour and generic information captured on the company’s website, for example, it’s not going to have the context or knowledge to handle a query correctly. Very few customer service chat bots get the job done as a result.

Service bots

The good news is, we’re starting to see the introduction of different types of chatbots that are better suited to different roles. For example, in the customer service space, service bots are being programmed and rolled out that can talk to the customer in their context. They have the ability to apply all the rules and follow the process so an outcome can be reached. In essence, they’re trained to marry the structured nature of business with the unstructured nature of human customers.

Going forward, we’re going to see more chatbots that can adjust to your context, and marry your context and world to the rigid structure of the company’s internal world so that an outcome can be achieved that’s satisfactory to both parties.

This is the next phase of chatbots. They don’t randomly learn, they learn how to apply rules in a better way to different situations. It doesn’t self-learn, rather it learns how to apply rules and processes better to different customer contexts. It doesn’t try to learn different ways of doing things because it isn’t allowed to.

Service bots will enable businesses to deliver the customer service engagement they’ve been striving for, and deliver better levels of engagement. This will free up human experts to handle more complex queries and deliver more personalised service to the customers they do need to talk to.


Main image credit:


Ryan Falkenberg is co-CEO, CLEVVA. CLEVVA is a digital self-service automation company that builds and deploys digital experts that resolve customer requests, queries, issues and complaints via digital self-service channels, without the need to bring humans into the loop.



– Receive the Retailing Africa newsletter every Wednesday • Subscribe here.