Something that initially attracted me to Monzo is the idea that it’s a smart platform. The examples during launch for alerting you if your waterbill is higher than normal or automatically asking TfL for a refund if you forget to tap out are precisely the sort of proactive, data-driven customer service I want from a bank.
Using the app since the start of the year, I see lots of opportunities where proactive alerts, suggestions, or actions would be useful. For example, I routinely spend ~£1.50 on a bottle of water and a chocolate bar during weekdays around lunchtime and I tag it with the same comment each time. I’d love the app to ask if I want to apply this description as soon as the payment happens. Or if I do an online shop, I’d love for the app to ask how much the bill is expected to be as soon as the preauth happens so that it can reflect my likely balance and even guess at what the actual value will be based on past experience.
As the Monzo proposition has developed over the last year, there’s not been too much talk about this stuff. Looking at the roadmap, there’s nothing explicitly about this and much of the development discussion has (naturally) been about the creation of a bank.
Two questions, then:
Is this still a key part of what Monzo will hopefully become?
Fostering the lovely sense of community here, are there any suggestions as to other automated warnings/actions/prompts that could happen? What would we like Monzo to help us with?
Completely agree with Ben, this is something I was wondering about too.
Has Monzo made any Machine Learning hires or have any future plans for a dedicated team? I believe that virtual assistant style guidance will be an integral part of future banking and personal finance.
(Perhaps the assistant can be a ‘character’ called Monzo and inject some much needed meaning/personality into the name).
In addition to suggestions, I would like to see machine learning applied to automagically (sic) extracting and categorising data from receipts eg. supermarket receipts that can have multiple categories of spending.
EDIT:
Interesting TensorFlow use case - sorting cucumbers!
When it comes to providing data driven features my guess is that :mondo: will focus on features it can monetize i.e. forecasting that users will run out of cash by the end of the month & offering a short term overdraft (as Tom discusses in this video) or telling users that they could save money by using a different credit card provider like ClearScore or Cleo (see below).
So it’ll probably be developers who drive a lot of this type of innovation, leveraging access to user’s data through Mondo’s open API, to build apps.
In fact something similar to this
has already been attempted & a pretty sophisticated sounding app was built, to deal with some of the challenges that came up when retrieving the data from TFL’s website.
Another example is Cleo’s chat bot which will eventually recommend cheaper finance service providers. It’s been discussed by the community here -
The company’s in the very early stages of development and I think it’s a stretch to call it AI but it’s an interesting concept.
Obviously the challenge is that, as far as I’m aware, only :mondo: will have access to the complete database of user data to ‘learn’ from so it’ll be interesting to see how far they go down this route.
It is a pretty complex technology & one that requires specialized (expensive) talent so I’m not sure that they’ll invest that much in this area, just to add new features, unless they’ll differentiate :mondo: from competitors significantly.
Having said that there’s a big difference between true deep learning and building programs that can share insights or automate processes, based on data & I’m sure they’ll do plenty of the latter.
Isn’t there a concern the AI could get too smart and start judging us on our purchases? Maybe it sees that’s the 5th time this week we’ve been to the doughnut shop? Starts declining the transactions until we buy some sort of salad?
Companies have been judging us on our purchases for years. In the words of ex-Tesco CIO Mike McNamara, the supermarket giant has been analysing customer data for “donkey’s years”, keeping every scrap of purchase data collected through the Clubcard.
While Tesco-owned Dunnhumby, the company which powers this big data analysis, employs more than 2,000 people in 29 countries, selling information from a 40-terabyte database, to companies including Procter & Gamble, Coca-Cola and US retailer Kroger.
Other useful ideas of things Monzo can learn to do:
Actually remember patterns for categorising transactions. From my experience to date it simply applies the same category I used last time for the same merchant. But truth is sometimes I want to make an exception and label that cash withdrawal as Groceries but only once. It would be good to look at deeper history and takes trends into account.
Once the spending target are live it would be useful to get proactive alerts if it becomes clear after s certain transaction I’m likely to overspend in that category this month. Something like “Thst’s probably enough shopping for September”.
having read this story about a company called Bonsai, who are offering a service to enable developers that haven’t studied deep learning in depth, to adopt AI
& listened to this podcast (13:05 onwards)
and heard Marc Andreessen say that in the last 12 months, AI has become a lot more accessible - so junior students have started taking open source AI and using it in their projects & at a recent a16z hackathon, most developers incorporated some element of AI into their hacks.
My guess is these were Stanford students & obviously an a16z hackathon is going to attract some of the most talented programmers. But perhaps Monzo’s access to AI isn’t going to be quite limited as I first suggested
I’d be interested to hear the thoughts of the developers in our community like @RichardR, @billinghamj & @-removed- (apologies if I left anyone out), as well as any :mondo: developers of course, on this…
