Introducing Ada: the newest member of the Tandem family
What does AI mean for the future of Tandem Bank?
What does AI mean for the future of Tandem Bank?
We’re lucky to have an incredibly talented data team here at Tandem. Recently they’ve been working on our new AI system that will help offer our customers a truly personalised experience. You might have seen a few bits about it in the press from our CTO Noam, but we appreciate some of them are a little heavy on the tech-talk. So, last week I sat down with the team on the front line of the build to talk all things Ada and what it will mean for you, our customers.
Noam: I’m the CTO at Tandem Bank. I make sure that Tandem is a data driven company and ensure everyone in the organisation maximises the benefit and value that we can yield out of data and put this in our customers’ hands.
Henry: I lead the data insights team here that manages all the data scientists and data analysts. My responsibilities are to ensure that Ada works for the scientists and the analysts and provides customer value through the use cases that the team puts through them.
Sam: I’m a data scientist here. My focus with Ada has been to extract the insights from the system, whether that be to the customer or the business. Organising the data in a way that’s useful for us to be able to push into customers hands or into the hands of the business to see how that analytics can help push us forward. This includes how we can visualise and present the data back in a way that gives the most insight and helps decision making for the customer and the business.
Henry: Sam is the important translator in effect – connecting the business language to the data language.
Beth: I am a Data Scientist (Graduate) with a current focus on financial crime and specifically AML.
Henry: A self-learning platform that is designed to deliver insights to add value to our customers’ lives.
Sam: A self-learning system. We have lots of information about our customers, about the market, about the population of the country, we can use that info to not only reflect but also to look forward and make predictions to help our customers manage their money better – foresee what they’re not able to see themselves.
Henry: No system can self-learn from nothing. She will be fed with business knowledge through Sam and the team and on top of that there will be the self-learning layer that grows her knowledge base.
Henry: The biggest thing for me is being personal. I don’t think any bank has actually offered a personal service yet even though the word is thrown around a lot. Our goal here with Ada is that she learns from millions of customers and so the outputs and insights she puts out, each one will be personal to each customer – there will be no generic outputs.
Sam: To shadow that, it goes beyond just the financial products to how the person would use and experience the app. So even if a customer is new to us, we can tailor their experience with our bank – that’s all through the app. Ada can help our customers manage all aspects of their money.
Henry: Banks are product focused – their core is the products they’ve built. They refuse to ever edit them. We’re taking the product focus out and putting Ada in at the centre – Ada is essentially the customer insight so we can focus on customer goals – we’re not there to get them a fixed term saver, we’re there to help them afford a house, for example.
Beth: Ada will allow rapid improvement to customer journeys and experiences. This will allow us to build and improve every aspect of our product quickly and securely, so we continually help to solve customers’ money problems in new and exciting ways.
Henry: To help them with their money, whatever issues they have, you need to be able to predict their cashflow. You know your direct debits and bills – they’re fixed. What people struggle with are the variable things like your credit card bill and sometimes people don’t think about how much they’ll spend on coffee/lunches next month. If we can start to predict the discretionary spend, we can really help them and get ahead. They don’t want to get three weeks into a month and think ‘wow I am going to be short on cash’, if we can, at the start of each month, give them a little forecast, we’re seeing that you’re on track for overspending. Or you’re on track for underspending then let’s help you maximise that asset.
Sam: I think that’s the main thing for me as well. What customers find difficult with money at the moment is this segmentation of products. So, if you have a lot of money in your current account, it’s just sitting around doing nothing, we can potentially predict which fixed term saver will be good for them based on how they’ve spent in the past.
Henry: Cashflow is a river with a finite source. If you don’t manage the flow of that source your river will dry and your boat grinds to a halt and you can’t do anything at all. But if you are doing well and your river is rushing, you don’t want to waste the water that’s spewing over the sides of the river – you want to use it in the best way that you can.
Noam: People forget that 20-30 years ago banks were pretty good service. You would come into a branch, someone would sit with you and make you a coffee, you would talk and make personal decisions together. They would try to run billions of algorithms inside their human mind to try to predict your cashflow and affordability and to analyse your financial future. That was done by humans for decades. With artificial intelligence, the team is writing the algorithms into the system to build the new 2020 personal banker, which is Ada.
Beth: Being able to feedback user experiences and learning from them mean that our product offering will be continually improving. The limitation really is our own imagination! Which is why it’s so important to work with customers and SMEs throughout the business to make sure that ada can work for everyone. For example on the side of savings, an improved cashflow model will help Ada to predict how much you have to save, and where you can make savings in your day to day spending, making suggestions to customers, and helping people who think that they aren’t in a position to save to build up good habits without even thinking about it.
Henry: Wherever possible we will explain the decision that’s being presented to you, you won’t be just given a number. We’ll always try to break it down and show you the journey we took to get to that number.
Sam: One of the things you can say is that because we are a fully regulated bank, we have a lot of responsibility for data protection and for being able to audit the way that we use data. We’re building Ada to be extremely accountable from the very beginning.
Noam: I would add to this that every piece of data we enter into Ada is tagged with all the relevant regulations so a machine can only be as powerful as you allow it to be. Every process that happens with the data is being documented by a machine which is very well organised rather than many siloed opportunities which could make it harder to control the data.
Henry: The complexity is making it personal. It’s very simple to make a machine learning model that does a very specific task in isolation. It’s very hard to have a group of models that work together to surface you the best, personal decision.
Sam: Part of the complexity of it for me is making sure everyone is able to interact with it, and that comes in many different forms. So if we want to put things out to the business, we have third parties so all of that needs to be integrated in. If we want to put things out to the app, we need to work with the people who develop the app what it looks like and the engineers – they need to be able to get the insights from it to put them into the app for the customer. There are so many inputs and outputs to the whole thing that it makes a lot of it really complicated. In the development over the last few months we’ve spent a lot of time talking as a full team making sure that each of the people that needs to integrate with Ada has their requirements fulfilled.
Beth: Making sure that we are feeding her with the correct information so that she can make the right decisions with the right data. Because it’s a completely new system, we have to make choices about the design of every aspect, which is an amazing opportunity for innovation, but it is also crucial to build in a robust way, so that we can future proof the system. We won’t get it right first time, but we are learning as she does, and building out processes and conventions which are truly Tandem specific and work perfectly for our internal stakeholders and ultimately for our customers.
Noam: It was hard work definitely; it is a complex machine. But we have a very wide team working on it from research, development, business and customer aspects and all of Tandem as a company is continuously investing and building this machine to serve the customer. We could not copy & paste or download some tool that does it. It’s a lot of fun because we know the benefit this machine is going to bring.
Get Tandem for free on iPhone and Android.Download for iPhoneDownload for Android