American Express Agent Assist

UX/UI

 

Role

UX/UI Designer

Duration

3 MO

designers create proof of concept designs to present to stake-holders in order to fund new products…

I worked closely with Product Managers, Stake-Holders, Designers, and the Engineering team to create wireframes iterations and a prototype for product which uses generative artificial intelligence to assist Amex Colleagues answer customer questions. As well as, analyze the sentiments of those speaking during a call. The prototype was handed to Engineering to develop a functioning product that was presented to Stake-Holders for Production Organization Approval and funding.

 

Problem

Create a Generative AI product that assists Customer Care Professionals answering customer calls, for POA (Production Organization Approval,) to acquire funding and continue design for widespread company use

Beginning this project, I met with the Product Manager, Senior Designers, Engineers, and a Stake-holder, about the product goal and what it’s meant to be on a high fidelity, scalable, level. This product is intended to assist Customer Care Professionals while on calls with customers by anticipating customer needs and offering quick links to journeys. Anticipating customer needs will come from the transcript of the call happening in real time. Generative AI also tracks the customer’s emotional sentiment on the call. This Sentiment Analyzer feature detects how the customer is feeling, as well as, pin points abusive language during the call. After the call ends, the Agent Assist also serves as a call log with the call information, transcript, sentiment analyzer, and summary which can be edited by the CCP.

 

Process

1.

April 2023:
Understanding usability by CCPs, Created low-fidelity wireframes of different design directions for both the Agent Assist modal on ISP and the Call Insights page. Meeting with the same group of people with stake in the product periodically every week to review the design directions.

2.

May 2023:
After design direction was agreed upon, I began to create more high fidelity wireframes. It was difficult to stop designing new concepts but the engineers needed a “final” file to work off of while developing a functioning product to present to acquire funding.

3.

June 2023:
I created a prototype to present to the Engineers on the team who would build this product. The Engineers and I would meet often to review the functionality was working as intended and UI was as close to the prototype as possible.

4.

June 2023:
The Engineers finished developing the working product and presented Agent Assist to the Stakeholders as Proof of Concept.

 

Solution

Assisting CCPs answer customer questions and update account information faster with Generative AI and after-call information available to review in an Interactions Log.

Though there is a testing pool of CCPs at American Express who give real-time feedback on the products designed for their use, because this was a Proof of Concept project, it was not tested with them. When the product is funded, the design goes through more exploratory designs, and a direction is finalized. That final design will be tested with the CCPs. Above is my design exploration for Call Insights. These screens express the call as a horizontal timeline similar to calendar application user interfaces. Splitting the call minutes vertically and adding dialogue, intents, and call sentiment from duration of the call.

 

Outcome

The Production Organization Approval was successful! Our team acquired the funding we needed to continue working on Agent Assist and launching it for widespread company use.

American Express being one of the bigger company’s I’ve contracted for, I hadn’t had to make a proof of concept for production organization approval before. I was happy to have been the designer to help acquire funding for a product that I believe will really help both Customer Care Professional do their jobs with ease, and have customers inquiries handles quicker, as to not be on the phone for long. Though my contract has ended I hope to hear about how Agent Assist tests with the CCPs!