Without wishing to sound utopian, AI can make payments better for everyone in every way. From now to 2027, the AI payment sector is expected to see a compound annual growth rate of 25.3%, which indicates the sheer enthusiasm for it among businesses. The excitement is understandable, given the potential that it has for cost-saving, customer experience, and fraud prevention.
Here is a breakdown of the growing role of AI in payments, what it will mean for the sector, and how you can use it.
Fraud (and error) prevention
AI will generally be very fast and accurate when detecting anomalies or suspicious patterns, so it can stop fraudulent payments and accounts. Machine learning can examine large data sets and make continually improving pictures of ‘normal’ and assessments of a transaction’s risk level. As AI becomes more advanced, the risk of fraud decreases.
On the other side of the coin, AI can also reduce the risk of a ‘false positive’ in fraud detection. Anti-fraud systems have triggers and filters that flag a transaction as suspicious. They tend to be ‘velocity’, or the number of transactions from an IP address, ‘value’, when the amount of the transaction is unusual, or a ‘mismatch’ filter between shipping and billing address.
Just as all of those filters are sensible and logical, it’s also easy to see how they can be heavy-handed and clumsy; businesses lose 3% of their revenue every year to declined but legitimate transactions. Not only do you lose a sale that way, but you can also lose future business if you irritate a customer into avoiding you temporarily, or even forever. Better AI means less fraud, but also fewer instances of those false positives.
For a long time, more or better security has meant more friction and a poor customer experience, and businesses have had to make a trade-off. It’s an exciting development that the two can improve in tandem.
Another type of false positive is the ‘false decline’. The system may believe that the customer has insufficient funds for a purchase, when in fact they do have the money. Again, the irritation to the customer and the brand damage can be severe, so the appeal of smarter AI is clear.
However, not all customer-focused AI has to be preventing poor experiences—AI can focus on building good ones. Here are some areas for that.
When data on buying habits is available, the AI can make very intelligent suggestions for additional purchases, or market additional products that the shopper is likely to respond well to. There’s a lot of scope for upselling and cross-selling, so it’s very positive from a revenue point of view. There’s a balancing act between helpful and sinister. Customers like to be understood and to receive help, but many shoppers still feel nervous about the amount of data companies gather about them. However, purchase suggestions seem relatively innocuous — a message suggesting ‘you might also like these items’, is considerably less alarming than the idea that your mobile phone might be listening to your conversations to choose the adverts you see.
A bad chatbot is worse than no chatbot, but a good one is brilliant for businesses and customers. Without the expense of a customer service agent, the business can provide care and attention to the customer or a potential customer, and the customer can quickly find the answers they’re looking for, and even transition seamlessly into their purchase through the chatbot console.
Good chatbots get that way through Natural Language Processing (NLP) and machine learning, and they need human interaction to learn from, so it takes more strategy, insight and effort to make your chatbot work well, than simply buying a solution and bolting it on to your website.
Amazon Go (Amazon Fresh in the UK) is the only example of an operational smart store, but it seems inevitable that more types will appear.
In Amazon’s ‘just walk out’ model, customers enter a turnstile by scanning their Amazon app, take the items they want to buy, and walk out of the store. In-store, tech can tell what the customer has ‘bought’ and charges their Amazon account.
It takes a suite of highly sophisticated technology to make it work. The shelves can tell when something is taken, but can also distinguish between removal for purchase and removal for close inspection. Cameras and sensors can distinguish which customer is walking out with what items. There’s IoT, AI, and machine learning at play, among many other things.
The potential is enormous. The business can
- pay fewer staff
- use more floor space
- enjoy the greater turnover from the increase in transactions per hour
- almost totally reduce theft
- keep a real-time track of stock levels and demand
The customer of course gets to shop without queuing or physically paying, and when the concept is thoroughly proven and embedded the time-saving will easily draw people away from traditional till-based shopping, and smart shops will quickly become the norm.
Like with chatbots, a bad execution is infinitely worse than not trying it at all. You certainly never want to charge people too much or not enough, or make people pay for each other’s groceries. Additionally, the upfront cost (at least at this stage) is enormous. The technology is young, and patented, so development cost or licensing is going to be the major initial overhead before the store can enjoy the savings.
How to drive change with AI in payments
In payments, change has to be technically possible, commercially viable, and user-friendly. That triangle is not easy to draw, and it’s rare to find people who can do it. RPI’s team have deep roots in payments — we find the leaders, visionaries, and technicians who drive change in the market, and we place them in businesses like yours.
Get in touch today and you can make sure your organisation is at the forefront of AI in payments, delivering the change that your business and customers need.