AI in Financial Services: Your Future Bank Experience

In the realm of finance, AI (Artificial Intelligence) and Machine Learning are reshaping conventional practices. This post sheds light on current trends in Financial Services, offering insights into how your bank accounts, pensions, and insurance might be influenced by this transformative technology. Keep reading to explore the evolving landscape where AI decisions, chat bots, identity verification, fraud detection, and more play pivotal roles in shaping the future of financial interactions.

Disclaimer: This isn’t an exhaustive list – it’s just what I’ve been talking to people about. There are hundreds of other examples! If you’ve got an interesting one please add in the comments.

Primer: What is AI?

AI (Artificial Intelligence) usually refers to applications where you combine machine learning or deep learning with a decision-making capability. In other words, you’re applying machine learning in your business processes, not just experimenting with it.

As with all AI, the hard work is all in the training and testing, not in the technology itself. The types of models best suited to each of these cases is well understood.

Here are some of the best examples of AI we’re seeing in Financial Services at the moment.

AI - Decisions and Actions based on Machine Learning and Deep Learning

Chat Bots

You’ve all used a chat window, right? These are increasingly being automated using AI, so that you chat to an application rather than a person. They’re available 24×7 and they respond fast!

Many organisations are starting to embrace chat bots for internal-facing processes (“I would like to book holiday”) where you can serve large numbers of people efficiently. 

Many businesses using these miss a trick by pretending that the chat bot is a person. If you let people know that the chat bot is a machine they’ll usually be much more forgiving with it. They will expect the machine to be imperfect but to be available all the time and be fast. Also, the way in which you’d chat to a machine would be simpler than you might to a person, meaning that the AI has a better chance of getting things right.

Soon you’ll be using a chat bot instead of filling out a form to get your bank account!

Picture of a friendly chat bot

Identity Verification

The AI to identify people from a photo is very established. Your phone will even suggest the names of people in your photos, so using the camera on your phone to verify your identity looks like a no-brainer.

Here’s the rub: what happens if someone tries to identify themselves as you by photographing a photo?

This is where this gets interesting! The leading tech in this area uses voice and video to check whether it’s really you. This makes identity much more natural and usable than having to remember passwords and use card readers. Expect this to be coming to a banking app near you in the very near future.

Fraud Detection

How do you know a transaction is fraudulent? Banks have been using statistical techniques for fraud detection for a long time. Your credit card company will likely know the usual payment sizes you make, and when something out of the ordinary happens you might get a text or a phone call.

What’s the benefit of AI here, when this has been working for years? 

Well, fraud detection is all about accuracy. Obviously what we’re aiming for is that 100% of good transactions go through and 100% of fraudulent ones are blocked. In practice it’s not as easy. We want to prevent as many false negatives as we can (transactions that are bad that the technology thinks are OK). This results in more caution, which means good transactions being flagged as potentially fraudulent. This is annoying for your bank (time and cost) and it’s annoying for you, because your money might be in limbo while the transaction is being checked.

The trouble here is that someone – a worker – often needs to go through the lists of quarantined transactions and verify them manually. 

Machine Learning and AI give us the ability to continually improve the accuracy of the detection process as we can feed the results of the manual checks back into the model to achieve better and better accuracy. 

Fraud detection - getting it correct versus false positives and false negatives

Cognitive Search

Cognitive Search is a technique that uses AI to improve search results, by enhancing the results and filtering them based on who is searching. This is a massive area of Data Engineering, and we’re only just starting to see some of the possibilities here. 

Some examples of what’s being done with cognitive search:

  • Finding a video based on spoken words within the video
  • Finding a document based on the contents of a photo within it
  • Translating documents as they are indexed so they are available in multiple languages
  • Understanding the purpose of documents based on NLP (Natural Language Processing
  • Understanding the intent of search terms instead of simply matching words


This is mostly being used inside big companies, to help make best use of their vast troves of digital documents. 

Data Protection and GDPR

You can use Machine Learning techniques to identify the elements within your datasets that are PII – Personally Identifiable Information. You can also be using this to match up the information relates to a specific person. This makes the process of compliance with GDPR much easier.

You can ask the question such as “What information do I have on person X?”, and you can get usable results. The requirements to be able to find the information you have on someone, especially when this can be email, voice recordings, letters and website/app logs means that it can be really hard to comply, even if you’re giving your best efforts.

Expect to see a load of AI-driven tools coming along to help in this area.

Trading Algorithms

I’ve left this until last, even though these have probably had more investment that all the others combined. The reason it’s last is that this is the secretive end of AI. The financial companies that are creating trading bots keep the contents of them very secret. 

Trading algorithm: using AI to determine when to buy and sell

In fact, the way that these bots are built are not that difficult at all. They’re trained to spot patterns in the price movements on financial markets and to make trading predictions based on these.  

There are downsides to these bots as well – when most of the market uses bots to trade, what’s really driving the prices in the market? Bots are usually followers of prices, unable to analyse and determine fair value. Despite this, the proportion of financial trades made by bots is huge. You can probably expect your pension to be under the management of a bot even if you don’t realise it!

Signing off...

The field of AI and Machine Learning is exploding at the moment, with new possibilities all the time.

If you’ve developed a great AI application and want to shout about it, or you have a great idea for how this could be used inside your company, please get in touch with me – I’d love to chat about it!

Share this post