NEIL PARKHURST

Omnichannel for Customer Service allows us to connect with a Virtual Agent. It is then possible to trigger a transfer to a human agent from within our topics / authoring canvas. But what if your customers are getting frustrated with a BOT that doesn’t understand them? In this post I will explain how we can use the “system fallback” setting to help with that scenario.

It may be common to construct a topic that allows the virtual agent to transfer the customer to a human. Maybe they enter an answer of “unsure” to a question and this could trigger the BOT to transfer to a human. (As I have shown below.) Or maybe you can have a topic of “human” that will transfer to a human if the customer enters certain key phrases. Like “I’d like to speak to a person” or “Can a human help me” etc etc.

But what if your BOT repeatedly does not understand what the customer is saying. (Or the customer doesn’t understand what the virtual agent is asking!) If you have ever experienced a BOT repeatedly saying “I don’t understand” you might know how frustrating this can be!

Imagine your virtual agent is asking the customer which colour they want. If the only options are “Red” and “Green” …. a customer asking for “Black” isn’t going to work. As an example I created exactly that scenario. Below you can see I am prompting for a colour and I have only added options of “Green” and “Red”.

Below is an example of the frustrating type of conversation which can occur. This is horrible as for the customer the loop will go on and on. Resulting in an upset customer. By the time they finally call to talk to a real agent they will already be annoyed. The BOT wants the customer to enter “Green” or “Red”. But this customer wants “Black”. The BOT tries hard by repeatedly offering an option it thinks is ok. But Red just isn’t want I want!

Note: This is just one example of many situations which could cause this type of problem. We could (and should) argue that your customers shouldn’t get stuck like this with a well-designed BOT! But in the real world we all know dead ends like this can (and will) occur.

Enable System Fallback

Luckily we do have a solution available to us! That being that we can enable a fallback topic which will be triggered if the BOT repeatedly asks the same question!

Below you can see that I can access the “System fallback” option from the settings cog on my virtual agent.

No we can add a fallback that will be triggered if the customer’s intent is unclear after two attempts.

You can see below that having clicked “+Add” a fallback topic has been created. I can now use the “Go to fallback topic” option to tailor it if required.

Below you can see the default fallback topic. Which you may want to tailor!

Now when I run the same test again you can see that eventually the virtual agent will transfer the customer to a human. Maybe that is still slightly frustrating but at least the customer isn’t stuck in a deadend.

I do advise you to try and avoid these “dead ends” in the first place. A design which avoids the situation in the first place has got to be the best option. But having a fallback for when everything fails also seems sensible to me. Enjoy!