Digital strategy
Mizuho Bank accelerates innovation in global transaction banking
Learn more
Homepage zuehlke.com
Banking
first direct launches AI autonomous banking to serve real-time customer needs
How Zühlke helped first direct introduce AI-powered autonomous banking to anticipate and respond to their customers’ needs in real-time.
-
Personalised banking experience helps customers manage their money better
-
Data harnessed to anticipate and respond to customer needs
-
first direct reaps the rewards of becoming a data-driven bank
The challenge
first direct launched in 1989 as the original challenger bank. Banking had been staid and steady for years when first direct adopted a banking epiphany: let customers do their banking in a way that fits around their lives.
To deliver this, first direct invested heavily in people, training, tools, and culture. And it worked. They garnered a loyal customer base of 1.5m, and regularly out-ranked their competition for customer satisfaction.
But over the past few years, mobile began accounting for an increasing amount of overall customer transactions (it currently sits at 84%) and first direct was being usurped by a new generation of challengers.
first direct needed to meet its challengers head on. They needed to move from using customer data in a linear, one-dimensional way, to being a data-driven organisation, anticipating and responding to customer needs in a way that excelled beyond their competition and provided customers with a personalised experience.
What we did
first direct turned to Zühlke to build out features utilising a blend of automation, artificial intelligence (AI), and machine learning, to anticipate and respond to customers’ needs and harness the power of autonomous banking.
The ultimate goal was to make it easier for new and existing first direct customers to intelligently and effectively manage their money.
Conducting a discovery for autonomous banking
The discovery process revealed there was huge value in creating solutions for customers to relieve them of the time-consuming, repetitive, administrative tasks related to managing their money. And many of these tasks could be automated or predicted.
first direct came to Zühlke with five levels of autonomous banking as their vision:
- Level 1 offers customers some basic reminders and prompts that a payment or action is due.
- Level 2 delivers intelligence around checking available balances to suggest payments to cover upcoming debits.
- Level 3 pushes on to monitoring bills and expenditure and making automatic payments where funds are available in the customer's current account and escalating to the customer as appropriate.
- Level 4 offers full 'self-driving' under certain conditions, making budgeting decisions and moving money between accounts to make payments
- Level 5 is full 'self-driving' in all conditions, running the customer's finances to an optimised position based on behavioural and attitudinal profiling.
To deliver first direct’s vision of autonomous banking, there were two objectives:
- To help communicate the idea of autonomous banking internally.
- Deliver a fully operational ‘feedback loop’ from end to end.
“The challenge was taking what we know about customers and their behaviour to make dealing with routine requirements straightforward and make anticipating likely needs a real-time reality.”
Building new data pipelines
The Zühlke team began by helping first direct understand what data it already had and build data pipelines with access to high-velocity behavioural and financial data.
The key was being able to use this data to react and interact in real time. Gradually, learnings began coming in for every interaction – or lack of action. For example, an action might be initiated by a change in how often a customer checks their balance. This meant we could start engineering relevant, personalised experiences for customers through the app.
Building Autopilot: a first in autonomous banking
The work and insights were then used to create Autopilot. This AI-powered function worked in the background of the app to enable and support customers at a level they wanted, for example, by making personalised recommendations and automating activities such as topping up savings accounts.
Creating feedback loops to support Autopilot
Zühlke identified how first direct could use the power of feedback loops to deliver and optimise Autopilot. We deployed feedback loops to take the AI model’s predicted outputs and reuse them to train new versions of the model. This meant we could reinforce the model’s training and keep it improving over time, reducing the likelihood of the AI results stagnating.
We used feedback loops to great success, for example to create a proxy that could detect stress in customers’ behaviour so we could then suggest steps to help mitigate their stresses, where possible. Along with this, we combined behavioural data with data from user research to test push notifications and messaging so we could iterate and optimise.
The results
We delivered our proof of concept in nine months and improved first direct’s existing mobile app by using real-time data to reach Level 3 in banking autonomy. This removed friction in the digital experience and intelligently automated banking tasks for customers.
With Zühlke’s support, first direct has continued to live the ethos first founded in 1989: reinterpreting customer service for the next generation.
Perhaps most importantly, trusting in the Zühlke team has given first direct confidence that they’re on the correct path. Evolving digital banking and customer service is a strategically important initiative that involves significant change and the potential for internal disruption. The proof of concept has been a real success, getting strong internal buy-in for the approach and its roll-out.
