Ava is Thomson Reuters first customer service chatbot pilot for lawyers. The design process faced daunting obstacles, including internal resistance to automation, but the outcome was worth it.

Ava answers questions about legal content lawyers have access to on Westlaw (the company’s legal search tool), expiration dates, account balances and payment deadlines. Instantly, the virtual assistant redirects users to requested online content, updates contact details and displays invoices through the chat window, keeping users in the context of their work.

The Challenge

A corporate-wide challenge questioned how chatbots could improve how the company serves its customers. It required identifying a use case and a target customer, as well as researching the perceived internal and customer barriers to automation. The use case and insights would be pitched to the CEO and the Innovation Office for potential funding.

My design team and I won the challenge.

A problem-solving approach

I chose to apply a Design Thinking framework to fit the ambiguous nature of the challenge, gain deep user empathy and explore lots of ideas. I planned a design roadmap of 4 design sprints over 2 months for research and ideation, that would conclude into rich insights on customer and business attitudes towards virtual assistants, and a viable solution. This Design roadmap later evolved into a Product roadmap for a pilot development and trial launch to a subset of customers.

Using my relationships across functions, I found our project sponsor in the Head of Customer Success for the Legal business, directing our choice of target customers to attorneys and small law firms who use Thomson Reuters flagship Legal search product, Westlaw.

How can we support lawyers as they purchase and use Westlaw?

The Westlaw product team was already looking for innovative ways to support these customers as they purchase and use the tool online. This became our starting point for research.

Customer behaviors and attitudes

Along with the senior UX researcher, I interviewed customers and internal functions.

A total of 28 attorney customers were interviewed one on one over the course of the project to understand their sentiment vis-a-vis chatbots and where painpoints arose in the course of their online purchase and use of Westlaw. Internally, both Technical Support and Customer Service were solicited to share where the painpoints were in their operations, and what were the most frequent and costly phone calls they answered from customers.

Key findings from customers:

• Lack of product usage history to inform the license renewal process

• Calling support or customer service is cumbersome and takes time away from working

• Login always causes problems (lost password, username)

• Login issues are the top and most costly calls within Tech Support

• Subscription-related questions (read “what do I have access to?”) are among the most frequent within Customer Service

• Account management calls are also most expensive within C.S.

Feedback from internal Customer Service:

This important comment from Barbara was a key opportunity for an automated solution to support both the team and the customer. This led to digging deeper into the most common and repetitive questions customers call about, the operational cost of these calls, and designing related conversations between a chatbot and a customer.

Ideation & Prototyping

We brainstormed all the above painpoints and identified where the biggest impact was. We explored the human language by mapping out various patterns of human-to-machine dialogue and patterns. Finally, we mapped out all the conversation paths for a customer service chatbot, based on the most common questions customers call agents about.

4 design sprints leading to a viable solution: a chatbot that automates simple and repetitive customer service tasks

We ran 4 design sprints, each exploring a unique solution that was prototyped and tested. This cycle of experimentation guided us to a final and viable solution: Ava, a customer service virtual assistant that answers attorneys’ questions on their subscription and account management in Westlaw. Technically, the chatbot capability communicates to an SAP database that contains customers’ subscriptions and account details, retrieves data to answer questions, uploads documents like invoices, and updates the data on the database.

Design

We designed over 60 iterations of conversational flows throughout the 4 solutions that we explored, eventually working closely with the customer service team to craft human dialogs based on the most frequent questions customers contact C.S. about.

Our design activities included:

• Identifying the chatbot personality

• Defining conversational design patterns to enable transactions and answers

• Stitching together multiple flows into a coherent experience

• Defining tone, design language and performance standards

Pilot Launch

Customer Adoption, Operational Efficiency, Cost and Time Saving

Ava saved $500,000 in operational cost and 30% of Customer Service agents time in the Legal business the first year, forecasting at least 25% increase in savings the following year as more customers adopt the feature. Another benefit to be seen was the positive result of customer service agents increased attendance to strategic customer issues.

Customers saved valuable time (and money) in getting answers that no longer required calling a human agent. On the business side, customer service agents were able to focus their time on more critical customer inquiries, leading to fewer customer calls and higher service efficiency.

Finally, we handed the designs over to the engineering team that subsequently built and launched a pilot to 200 beta users of Thomson Reuters WestLaw.

Bonus: We published conversational UI best practices!