Demystifying AI: Q&A with Guillermo Ortiz de Zarate

Written by Guillermo Ortiz de Zarate on February 27, 2018

Editor’s note: Associations everywhere are using AI to make their processes more efficient. We’ve all heard by now of the radical potential of AI, but it can be hard to connect the dots from these big ideas to small practical steps. We want to demystify AI to make it accessible to techies and Luddites alike!

Recently, we heard about a new tool being deployed by NCARB to automate their customer service interactions. Zendesk is a software designed to provide fast and efficient customer service using an evolving knowledge base. We posed some quickfire questions to Chief Information and Innovation Officer, Guillermo Ortiz de Zarate, to get to the bottom of how NCARB implementing AI, what drove them to do it, and what are the expected results.

How does this AI technology work?

“The technology uses a Knowledge Base (KB) which you maintain by writing articles that answer common questions, like FAQs. You can use this KB to answer manually by linking to the article in your response. Or through taxonomies, you can set the system to automatically answer the question straight away. The system chooses a few articles that look like they might answer the question and sends it to the user, asking them if the articles helped answer the question. If they click yes, the machine learning algorithm takes note of the success and uses it in the next iteration. We receive thousands of simple questions through our website that could be easily automated, so we are working on the KB articles right now.”

How do you expect the implementation of ZenDesk improved efficiency?

“Probably over 90 percent of the questions that come to us through our website are repetitive and super simple. The problem is the 10 percent that are super complicated. We prefer to spend time taking care of those and let the system automate the others.”

Have you come up against any barriers so far?

“One barrier has been improving our Knowledge Base. The articles must be good enough that we trust the system to send them automatically. The other barrier has been the anxiety that automation causes some team members. It’s hard for people that are accountable for the customer experience to let go of that responsibility, even if it means less work. It is a process.”

What made you choose ZenDesk?

“We were working with Zendesk in our IT Helpdesk department to begin with. We started working on migrating the entire Customer Service department to the same tool when we learned they were launching the answer bot.”

What advice would you give other associations looking to adopt AI-powered tools?

“You don’t have to adopt AI across the board all at once. Maybe you start out by automating some of the smaller channels of communication. Then you can measure a couple of things, like the impact it has on efficiencies and effectiveness in providing customer service, and the impact on sentiment. Do users feel the experience is better or worse? Use that data to inform how you migrate other queues or touch points.”