AMERICAN EXPRESS CASE STUDY

CHALLENGE

To create a new way to share meaningful credit card referrals

Using a wizard-of-oz method, we rapidly iterated a chatbot to engage users to find a card to bring users value

OUTCOME

Designed conversational AI and design, conducted and facilitated user research, presented to key stakeholders

ROLE

The Challenge

Amex wanted to make it easy for cardmembers to refer friends. They thought a chatbot might be a good solution, but there were many unanswered questions: how can we build this? What existing internal services does this integrate with? How might we create a good customer experience? And of course, will it even move the needle?

AMEX Experiment Hypothesis Dashboard

Over the course of 16 weeks, we created a variety of functional and static design prototypes to test 25 hypotheses with 50 target users over 9 experiments.

What we did

Philosophie facilitated a workshop with stakeholders to identify the concept’s underlying assumptions and reframe them as testable hypotheses. We prioritized the riskiest assumptions and established a weekly prototype-and-test cadence to systematically validate or invalidate them.

AMEX screens

Key research insight

The key insight from our design experimentation was that customers preferred a multi-channel approach as opposed to only interfacing with chat. The chatbot was not perceived to be secure enough to discuss credit card details. As a result, the ultimate design solution supported a seamless customer experience across text, web, and existing customer service channels.

What we made

Several POC iterations of a chatbot for prospective customers to learn more about cards and apply. We created several conversational UI scripts with tree branching and defined HITL (Human In The Loop) touch points. The key was to create an engaging prospect experience that leads to a relevant, personalized and beneficial credit card recommendation.

Services

Product strategy, user research, UI scripts, customer journey map, user flows, rapid prototyping

Technologies

Twilio, Google Voice, Sketch, InVision 

The Result

  • A Wizard of Oz prototype to test relevant tree branching

  • An Amex branded Twilio prototype for stakeholder socialization

  • A detailed hypothesis dashboard to track every experiment, visualize validation, and gather research feedback

  • Created a customer journey to show the key touch points and chatbot UI recommendations

  • Qualitative user tests with 38 potential customers and 12 stakeholders

AMEX logo

 “It got to a really great place. It’s the right level of traction to move it forward.”

– Carl Barkey, AMEX Vice President Enterprise Digital Acquisition Products