There was a lot of buzz last week when Facebook rolled out suggestions from M–its AI-powered virtual assistant–to all Messenger users in the US.

 

 

Chatbots are fast and easy to develop

As most of you know, Facebook has been actively promoting development of M-style chatbots for messengers for a while now. A year ago they sent out a post to the developer community letting people know how to do it. And last fall a report from Facebook executives put the number of chatbots in use at 34,000. That number is surely much higher by now.

As we’ve been discovering, integrating natural language / learning chatbots into a Customer Experience Roadmap is a hot topic. Customers want self-service. And customer service managers welcome new ways to deflect easy questions and get more information on customers before bringing on a live agent.

Now, before we weigh in on the question in our blog’s title, let us answer another question that’s probably in your mind. Why would you care about our views on how to evaluate optimal AI use for your CX organization?

Here are three reasons:

1.Our founder was part of the first team in the world to publish research on neural networks for language models working within dialog systems (2000 ICSLP).

2.  We are the only B2B chatbot vendor on Bloomberg’s Top 50 Most Promising Startups in 2017.

3.  We are in pilot programs with Fortune 1000 corporations to help revolutionize how AI can help them deflect live contacts, improve FCR and overall CX metrics.

Now, as to the Facebook Messenger capabilities.

It does have some “intent” and learning capabilities, and this new development will provide value to messenger uses with things like offering up Uber rides, or providing payment information to save you a click to another app (very clever, Facebook).

However, for those of you trying to provide support for more complex interactions, you will need a system that is designed specifically for your needs. It should also be configurable (ie, something you don’t need to code). In short, your chatbot should really be a true virtual assistant – capable of accessing all your data (FAQ, ZenDesk, CRM) and proactively offering data that can divert the need for a live agent. It should also provide customers with a very satisfying self-serve experience.

So, while a M chatbot might be quick to build and capable of handling a single task, the odds are that you don’t live in a single answer relationship with your customers.

Therefore, my advice: think bigger.

With that in mind, we’re kicking off a blog series called “5 Questions CX Managers Should Ask Before Building an AI Chatbot.”

In the series, we’ll examine 5 key considerations that our CX partners have asked:

  1. How much will it cost and is my ROI aligned with the solution provider?
  2. How long will it take to roll it out?
  3. What are best technologies to invest in?
  4. Will it offer me real-time controls over my brand?
  5. How many and what type of use cases will it support?

Of course, we have a viewpoint on each of these topics. We’ll also be sharing recent feedback from leading corporations, which are doing pilots with us. If you sign up for our newsletter, (see the box in the right hand column) you’ll get all of this information once the blogs go live.

Lastly, we want to mention our solution brief, which provides a quick overview of our product and vision. Download it below to learn more.