Fusing emerging consumer AI with human-centred CX to inform ongoing innovation
Westpac - AI-driven innovation experiments
Over multiple years, we partnered with Westpac Australia on AI-driven proof of concepts for customers and internal staff. Together, we explored emergent conversational AI and multilingual automated voice prototypes in collaboration with Google's Translate team, through to multi-modal conversational UI for young adults looking to open their first bank account.
At a glance
The AI-driven proof-of-concept initiatives were conducted in 6-week sprints.
Key outcomes from these experiments include:
Identifying use cases where AI and digital humans can provide value that real humans can't - especially in sensitive situations. Users can feel judged when sharing their circumstances with humans compared to digital humans.
Recognition of mistrust and vulnerabilities associated with automated voice solutions.
A process to develop robust, dynamic knowledge graphs or knowledge bases from validated use cases, that can then be delivered through multiple human-centred interface modes.
No items found.
Rather than focusing on immediate business outcomes, Westpac aimed to assess the feasibility of ideas for ongoing investment.
Use case 1 —Wendy the job coach
Thanks to the results of our proof of concept just 6 weeks from concept to launch, Westpac were able to prove that there was genuine potential in ‘Wendy the digital job coach’.
Key learnings from testing and optimisation with real users:
1. Seamless interface modes
Seamless switching between different interface modes based on users' environment, context, and specific needs was crucial.
For example, Wendy should sometimes read her lines verbatim, while other times she should move to the background, allowing users to explore longer-form articles or video content at their own pace.
Example idea from user testing:
2. Push-to-talk
Initially, Wendy’s digital human interface required the microphone to always be on so users could go hands-free.
However, user testing in noisy environments revealed how often Wendy inadvertently “listened” to people talking in the background, replying nonsensically and frustrating the main user. Testing participants sketched up ideas around a way to ‘interrupt’ or pause Wendy:
After further ideation, a ‘push-to-talk’ button was introduced. Although not hands-free as per the original vision, users in any environment were now able to control exactly what utterances became part of their dialog with Wendy.
3. Don’t launch early
Our research revealed that in general, conversational interfaces (eg: chatbots) were being deployed prematurely, offering little value to users, and causing immediate friction and negative brand perception.
Many of these solutions expected users to restart conversations as they switched modes, or interrupted and then resumed conversation streams as a connection dipped in and out
This meant that conversations lacked awareness of the customer's history or context with the organisation, let alone the previous parts of today’s conversation.
In emerging technologies, there aren’t well-established patterns and best practices to fall back on. Testing with a range of users in different environments, before rushing ahead with launch, is crucial to ensuring the experience delights rather than frustrates.
Why was Wendy given the role of ‘digital job coach’?
Westpac aimed to engage the next wave of young people by helping them in the hunt for their first job. This is also a time when people may be opening their first bank account in order to be paid. The hypothesis was that thanks to engagement with Wendy, Westpac could be front of mind.
“Data released by Westpac in 2020 showed almost a third of young Australians aged 14 to 18 are not confident tackling the initial steps of searching for a job, in particular navigating finances for the first time, with more than two thirds of teens (71%) not feeling confident opening a bank account on their own to get paid.”
No items found.
Use case 2 — Training call centre staff
A second proof of concept was also launched, this time internally to train call centre staff.
There is a huge burden on call centre leads to effectively train new staff members in the correct etiquette and tone to speak to customers in. While this training can be particularly nuanced when customers call in with complaints, there are some textbook do’s and don’ts that new staff simply need to practice. Simply reading the rules isn’t enough; new staff need to practice speaking, receive feedback, and practice again.
We partnered again with the innovation team to trial a concept for new call centre staff to be trained by Wendy in early customer conversation etiquette. The qual feedback from staff who tried it was promising; however Westpac decided not to develop this concept further.
No items found.
Some of the technology behind Wendy
Some of the emerging AI technologies Wendy leveraged were:
Google Cloud Services, including STT / TTS (speech to text, text to speech)
The autonomously animating avatar which was able to react in real time to the written, verbal, and facial sentiment of users was created by Mark Sagar, the Oscar winning pioneer of computational modelling of the human face, and his start-up Soul Machines.
No items found.
The team at Ghost Partners have been helping us to launch emerging technology with a strong focus on customer experience.
Working across multiple time zones as extended members of our innovation team the quality of their work, honesty, and collaboration are more than enough reason for me to recommend them to any organisation – big or small.
