Artificial Intelligence in the Contact Center

Centerfirst has performed hundreds of thousands of call and case monitors for healthcare contact centers and is continually looking for ways to bring today’s monitoring automation to our clients. Today we are speaking with Chris Martinez, CEO and Founder of Idiomatic, Inc. to learn about how he thinks about Artificial Intelligence and how Idiomatic is being used within contact centers in some of the most innovative companies in the world.

Artificial Intelligence

Chris – please describe what Idiomatic does and how it came to be… 

Idiomatic uses artificial intelligence to help customer service teams unlock the wealth of information they are collecting in their conversations with consumers. We mine conversations using Machine Learning to automatically understand and categorize free-form feedback at scale. We uncover human-understandable insights directly from text so marketing and product development teams can get exactly what they need without the customer service team having to do any manual analysis.”

 

I have heard you speak about the contact center customer interaction in terms of Contact>Route>Diagnose>Code>Resolve>Respond. Please share how you came up with this call flow model and how your customers have used Idiomatic to make this an improved and more efficient experience for their customers.

“A lot of what we’ve come up with has been developed just by being in the field and talking to call centers and agents about how they currently do their work. We can’t take credit for inventing this flow, but we do have something to say about how to make it better.

Within the flow of Contact>Route>Diagnose>Code>Resolve>Respond, there are many areas where our customers improve using the insights Idiomatic provides. The first step is to have Idiomatic automatically Code the cases with very granular, actionable categories.

Once you have cases being coded in this way you can slice your data to improve the entire process.

One of our clients wanted to improve Responses. They could see precisely which codes are leading to negative customer satisfaction scores and zero in on where they should prioritize their efforts to improve the quality of the responses that agents have available.

One way to think about it is because we automatically code cases that are coming in this week, our customers can focus their improvement efforts on the areas across their workflows that most need it to help handle the future cases that you encounter next week.” 

 

Most of your clients are in the technology space, but the exact same customers have the same expectations when calling a healthcare or bio-pharma company. What would you see as a natural place for us participating in the healthcare space to begin or make advancements in using AI?

“We split the world into two types of benefits AI can bring: insights and improved workflows. Insights means actionable intelligence to answer the question of what exactly are your customers experiencing and/or struggling with. Normally insights are being communicated out to the broader organization (like marketing, product development, brand, etc.). Improved workflows means getting better at handling incoming customer service cases.

To us, the starting place for both has always been about the automatic human-like understanding of cases through the form of automatic coding. The first immediate benefit you get from that is insight into the customer experience. Our customers can effortlessly answer questions like:

  • What are the top 3 things we can change to reduce the number of questions we will get next month?
  • What are the new issues experienced because of our recent product launch ranked by frequency?

This precise coding empowers and informs everything else you try to do on the workflow side. If you want to have automatic answers, it’s probably just on a certain category of cases (so you have to have them categorized). If you want to improve net promoter score , you need to know not just that NPS went down but which types of issues increased (so you need them categorized). Almost anything you can think of doing boils down to having a fine grain categorization of cases as a baseline.

So, what we’d say is start with insights, see if you can find some system for helping you better understand and communicate what’s happening back to your company. Once you have this better understanding then you can decide where to invest your time, money, or effort to come up with improvements (or 3rd party solutions) to help you solve the workflow problems that matter.”

 

What are the limitations of AI and what do you see as the time horizon for overcoming some of these limitations?

There are many limitations to A.I. at the moment. At a high-level we can include:

  1. Training Data
  2. Simulating human-like emotion (such as empathy)
  3. Context

Obtaining large, clean data sets for the training of Machine Learning models can be a very difficult task. Some of the best algorithms these days need lots and lots of data. You see folks like Google and Amazon building the cloud infrastructure that allows large datasets to be hosted remotely and used for training models. This is getting better as we speak.

Simulating human-like emotion and empathy is a very difficult challenge for A.I. in all but the most simple of cases. We can train a narrow chatbot to answer a very narrow set of questions (almost like canned responses) but having one understand sarcasm or humor is incredibly difficult. Something like general positive, negative, neutral tone/sentiment is more in the middle. For things like chat and emails we see this changing in the next couple years where technology will allow us to simulate human understanding on more and more complex issues. For phone / voice we’re still pretty far away. Just go tell Siri a joke and you’ll see.

Finally, and perhaps the most difficult challenge is imbuing machines with the context that we develop just by being alive and walking around earth for a few decades. Having the context to know something like when “water under a bridge” means water literally and when it is being used as an idiom is extremely difficult. This is the challenge we work on every day, hence the name of our company “Idiomatic”. Personally, I think we’re decades away from solving this problem.”

 

Is there anything else you would like to add about Idiomatic or AI?

We are really excited to be working in this space. Imagine a world in which every customer is heard and companies can actually respond and fix the root cause of customer issues. No more customer service run-around, getting automatic responses, or having to call again and again and again. Needless to say, we think of it as a much better world. Idiomatic is working hard to enable our clients to bring about exactly this type of experience for customers.”