3 qualifying questions

3 Qualifying Questions for Chatbots in Customer Service

Not all customer service organisations should invest in chatbots. Answer these 3 questions before even considering it.
It's Not For Everyone
We all know customers get annoyed when they have to wait in line. Whether it's waiting 12 minutes before a phone agent picks up their call, or 12 hours for a reply to their email, waiting just isn't fun.
There's a good reason why digital virtual agents are taking off in customer service organisations around the world: the promise to significantly reduce first response times (and ideally resolution times), is a tempting one.
However, if your customers are not reaching out to you through digital channels, or you have expert customers that aren't asking you standard requests, then you'll probably have a hard time making your chatbot project ROI positive.

Three Qualifying Questions

Before answering the type of chatbot you need, you should answer if you need one in the first place.


1. Do You Have High Inbound Support Volume?


If customers are not contacting your customer service organisation proactively and at scale, there's little need for a virtual agent. Chatbots are not (yet) great at reaching out to customers on their own, but should focus on inbound requests.

Ideally, you already have a significant volume of digital inbound requests (i.e. email, webform, or chat) and a clear page where customers go to contact you. This way you can easily influence your clients to give your virtual service agent a go instead.

If most of your clients are contacting you via phone, it's still possible to change their behaviour, but it will require more education (mentioning the virtual agent while they wait in line, sending an SMS with a link to the virtual agent, etc.).


2. Are Customer Requests Significantly Repetitive?


Are your customers experts in their field and extremely literate in your service? Then chances are they're not going to ask you the same questions en-mass, which means it's really hard to train a virtual agent to automatically resolve these requests. This is more likely in a B2B setting, but can also be the case for e.g. developer-facing services, like GitHub.

On the other hand, if you're targeting the mass consumer market (e.g. like a Bank, Telco, or Online Shop does), chances are high that you're dealing with a significant volume of repetitive requests. 

With our customers, we typically see that their customer support requests follow the 80/20 rule: 20% of problems account for 80% of the requests. In these cases, the right virtual agent can really shine!


3. Are You Able to Categorise Requests?


We say "can" above, because it's still not for granted that a chatbot will work for your organisation. Wether a chatbot is going to deliver the desired results, still depends on the type of requests your customers have.

 Support Request Type Matrix"Support Request Type" matrix


In the Support Request Type matrix above, you'll find two variables that significantly impacts the ability of chatbots to deliver value:

  • "Get-It-Done" vs "Troubleshooting" requests, and
  • "Self-Service" vs "Need-Agent" cases


Most chatbots are really only providing significant value in the get-it-done self service cases. These are the "how do I reset my password?", "can I return my order?", type of FAQ requests.

In order to solve the need agent cases, rather than handing over to a human agent, the chatbot vendor needs to enable you to retrieve the required information from your backend systems. A get-it-done need-agent case, might be "please block my stolen credit card", whereas a troubleshooting need agent case might be "what's wrong with my device?". 

In these cases, the chatbot need to be able to communicate with an API on your side. If the chatbot vendor cannot do this, or your organisation doesn't have the relevant APIs available, all the chatbot will be able to do is to help pre-qualify for your human agents.

Troubleshooting requests are problems where the customer typically cannot clearly articulate the problem they're facing, only the symptoms (e.g. "what's wrong with my device). 

If your customer service organisation is facing many troubleshooting requests, you need to evaluate whether a chatbot will be able to help your customers. Asking clarifying questions becomes key. Try out the "Helpbot" from Formlabs here for a good example (yes, they're a Solvemate customer ;)


Our CEO also wrote an article on this topic on Medium, so in case you want more details, check out his 3 Conditions For Successful Chatbot Implementations.

Did you find our post insightful?
Share it with others