User Research & product design
Using AI & machine learning to boost personalization
Backbase is an Engagement Banking Platform company supplying white label experiences to millions of users worldwide through partner banks.
My role
I was the sole UX Lead doing the research, problem defining & solutioning at a high level for the entire product.
The product
Digital Engage is a B2B SaaS application that allows marketers to communicate with the end user of the banks. They can do this via touchpoints such as banners, message centre, overlays & push notifications. Marketers use the app to upsell, cross sell and increase loyalty to the bank.
Background
The Head of AI joined Backbase and looked at places across Backbase where we could make impact. During this time, I was leading the assessment of opportunities with AI in our product and where we could make the most impact, both from a commercial standpoint and a usability standpoint.
The team
I worked with: Product Director, Head of AI, another UX Designer, Lead Product Manager, Development squads
In September 2024, I started the initial discovery of using AI in Digital Engage in order to make our platform smarter, more engaging and more efficient for our marketers & the end users of banks. My discovery focused on the marketers process and where we could increase personalisation in order to meet the vision of our product, enabling one to one marketing.
At the same time, the Head of AI joined Backbase - he was also conducting discovery on where we could add AI into our product suite.
Initial discovery
Initially, I was working internally on a journey map with specific opportunities based on friction points from previous user research. This led me to consider the following opportunities:
Audiences
A marketer often creates audiences from pre-defined segments - e.g. age = 18+; credit score = Good or Excellent. This takes time and effort, usually needing data analysts. But, it would be better if machine learning could take into account different parameters to build audiences and recommend them to the marketer.
Delivery of message
In our platform, marketers could not personalise or get any info on when is best to send to a group of users, let alone individual users. It would be better if the machine learning could analyse data individually to personalise the message time for every different user.
Content
Marketers need to go through a full process, sometimes with an external agency, to create content. It would be better if we could use GenAI to come up with messages that work, while learning from itself to keep optimising.
AI Channel
An end user can sometimes get a "banner blindness" type of psychology response to marketing. Maybe we could create a channel that is AI based, delivering marketing messages via this channel regularly.
Merging efforts
It was fairly clear that we & the Head of AI were duplicating efforts. In October 2024, we decided to merge efforts.
Discovery workshops
I hosted a discovery workshop where we went through the vision that the Head of AI had, customer problems, our ideas, journey mapping and building blocks.
Multi disciplinary workshop
In the workshop, we had engineering leads, the Head of AI, Product & UX to get a wide range of roles in the room in order to collaborate early and often.
Next steps
Once we had mapped this out, we had a clear problem statement: banks are struggling to keep customers active and increase the customer lifetime value. More and more people are moving banks as onboarding becomes easier. We want to increase that by bringing smart data into our marketing platform.
We then started designing.
Creating a baseline for the AI & ML
Create a product
In order for the AI to send products to users, we need to setup the products beforehand. For example, 5% savers account.
Select product benefits
To craft messages, the AI will need to know what the benefits of the product are - for example, you can earn up to 5.6% interest.
Generating 90 day plans
AI generates a 90 day activation
AI takes the baseline from the products above, takes into account the type of products you want to list and then generates a 90 day plan.
See each activity
Once generated, the marketer is able to see each activity, with each channel that is used. This puts the user in control of what happens in the 90 day plan.
Edit your 90 day plan
If the marketer needs to edit an activity, they are able to by clicking into it and selecting the channel - this will let them know key details & give editing rights.
See how it's performing
To understand how it is affecting your customer lifetime value, the marketer need to understand how the activities are performing - we give that initial info.
The app has not yet gone live, so I will instead speak about my learnings so far…
Greenfield projects rely on UX x Developer collab
I hosted weekly sessions with the developers doing the work so that we could align between UX and Development on the feature set. This stopped a lot of miscommunication and misunderstandings between us which sped up development time significantly.
Go quick & break things
Try out new patterns but make sure it goes quickly - especially with projects like this where we're uncertain on if it'll meet the solution will meet the success criteria we set out in discovery. Using the Design System, we can get it into the hands of our users quickly.