Esri’s sales chat:
This is Esri’s first chatbot. It was integrated to connect users to Sales, tech support, and customer service. It utilizes AI and natural language processing to query questions and get people to the right person to answer the right question.
Impact:
This work resulted in a decrease in misdirected sales by 50% and a decrease in dollars spent per day redirecting calls by more than half.
Context:
Esri needed a way to redirect users who were utilizing sales chat to answer tech support and customer service questions. This practice was cutting into Esri’s bottom line. We were spending $1,500 - $4500 per day.
Define:
I held a series of workshops, inviting members of digital sales, tech support, and customer service to help define what a chatbot means to Esri and how it could help Esri and their department.
These were business needs that we needed the chatbot to solve. We wanted to make sure it was branded correctly, in both how it looks and how it speaks. We also wanted to make sure it connected users to the right place.
Empathize: User Interview:
I conducted user interviews with 10 participants to understand what users’ current experience with Esri sales chat is and how we might improve it.
Why this matters:
8 of 10 users mentioned that they might as well call. Using chat was slow. It took just as much time to connect and get the answers, as calling did.
Validating this gave us the confirmation that our chat’s inefficiencies were not just a business problem, but a user problem, and worth fixing.
Outcome: Personas:
Personas were referenced in all subsequent defining meetings. They helped me communicate with stakeholders and were referenced when designing features.
This is our IT specialist persona. Their motivation is to get help from tech support with install instructions. We also had personas for a buyer who buys a product and a customer service persona who needs help with their account.
User journeys:
From the user interviews, I also created user journeys, showing how folks go about trying to get their needs met at Esri.
Dana’s Journey:
This is a user journey. It shows how Dana goes about connecting to Customer Service. The experience is ultimately frustrating for her, as there are 10 minute wait times before she is connected.
Conversation design:
I engaged with our writers and product owners to take a first pass at constructing a chat tree, to show how we could structure the conversation using the bot.
This is the first draft of the dialogue tree for the chatbot. Since speed was important, we focussed on trying to get users answers quickly.
Wireframes and testing:
I created a clickable prototype and put it in front of 6 users. It was a 45-minute test to validate user flows for the above dialogue tree, and to see if there was anything that we missed.
Findings + Iterations:
We discovered that people did not understand what was under Sales, Technical Support, or Customer Service. We needed to pull those options forward. Furthermore, people were confused with our navigation options.
Iteration:
This is an image of the revised menu options. We tested this again with a new set of users and had them try to solve them. They tested much better, and this gave us the confidence to move ahead with our designs.
Comps:
With basic flows out of the way, the product owner and I took a look at more unique flows. The example below shows how we showcase a link.
Connecting with an Agent:
We also wanted to differentiate the bot from an actual chat with a live person. We opted to use an icon, with the first initial of an agent’s name. I also thought it was a good idea to provide a subheadline that states when an agent joins the chat in real time.
Results:
After launching the bot, we noticed an immediate downturn in the amount of time our agents spent with a user. That time was cut in half. We also noticed an increase in traffic to customer service and tech support. Additionally, we noticed an increase in sales by 30%.
What I learned:
How to own a research process:
Lead internal workshops to get at critical questions
Drafted study materials
Launched tests
How to own a design process:
Dictate expectations and requirements from me, for chat
Which deliverables to be expected
When we needed to re-think a process
How to manage and resolve conflict:
Lead workshops on effective ways to communicate
Got alignment on where the documentation should live, and what everyone needs
Worked to define success criteria for a successful product launch
You got to the end!
Thanks for stopping by.
If you are interested in seeing what else I can do, drop me a line!