By the time you read this you may be tired of hearing about ChatGPT the OpenAI chatbot application or even its rival Bard, the experimental conversational AI service that has been introduced by Google. But whether or not these innovations have been over-hyped, there’s no doubt that they are finding wide application across multiple use cases. By Ralf Gladis, CEO, Computop .
Despite only being launched for public use at the end of November 2022, ChatGPT is already being used to create personalised, automated answers to customer inquiries across the eCommerce industry and marketing departments are relying on its language processing abilities to curate content for social media and online marketing campaigns. Students are also finding its capacity for crafting responses to their questions helpful when it comes to writing essays. Such is its popularity it attracted 13 million individual active users per day just two months after its launch.
ChatGPT is of course, the free, user-friendly frontend behind which sits complex technology known as generative AI. Generative AI is described by the World Economic Forum as ‘a category of AI algorithms that generate new outputs based on the data they have been trained on’. Like traditional AI, it uses neural networks that allow it to recognise patterns and make predictions, but generative AI also creates content in the form of images, text and audio. This significant progression means that in future we are likely to see the impact of generative AI across multiple industries, including payments.
Let’s have a look at how this might develop:
AI-enabled chatbots are already making their mark on social media platforms where they support sellers to deliver a vastly improved customer service that enables instant responses to customer inquiries. But there is no reason why these interactions couldn’t lead to transactions enabled through AI APIs. These could facilitate biometrics such as face recognition that, in turn, would connect to a payment mechanism such as ApplePay or GooglePay to complete the transaction. One thing that any merchant would have to ensure before using AI bots however, is that they behave well and are trained to be polite with customers at all times!
Companies that provide online or phone support to customers are already utilising AI to provide them with information about their customers to inform the conversation. Using generative AI, however, delivers information on a much wider range of topics and in much greater depth which would allow the AI to manage a bigger number of interactions freeing agents up to deal with more complex tasks. If customers then want to make a payment, as with social commerce, an AI user interface could use payment APIs that would enable payment through biometric authentication, or through a series of ‘if this/then’ conditional rules.
It is also exciting to think about AI behind personal assistants like Alexa or Siri. If a customer wants Siri to book event tickets or a hotel it requires machine-to-machine interfaces to automatically process the booking and the payments for customers. Obviously, there are many tasks that are given to a personal assistant and many hotels, ticket shops and retailers that consumers want to buy from. Providing APIs to handle all of this is a task that would require less AI but a lot more in terms of global standardisation and hard programming work.
AI is already having a significant impact on the payments industry by enabling a deeper understanding of customers and customer behaviour which is improving payments services. However, generative AI can take improvements a step further by processing more data at a faster rate which will help to provide real time insights, improve procedural efficiency and allow companies to meet regulatory compliance more easily.
One of the biggest contributors to credit card fraud in recent years has been the growth in eCommerce ‘Card Not Present’ transactions. Payment processors, like Computop, are already incorporating AI and machine learning to assess past data that can be used to help combat these fraud attempts. However, because generative AI is about creating content based on existing data, in helping to manage fraud, it would be best used to help construct messages to customers or new fraud rules, or through machine learning models, assess the performance of these rules to identify if they are working and boosting performance. If an AI model can reliably filter out low risk fraud alerts, manual reviewing efforts can be reduced, saving organisations time and resource costs. However, the challenge for AI in fraud prevention is to avoid biased results that might result from unbalanced training data leading to unfair judgement of skin colour, nationalities etc.
A frictionless checkout experience is the goal for all online retailers. This means that customers must be able to make payments quickly, securely and conveniently with minimal clicks. They are reliant on payment service providers to facilitate this, and the sector is now competing on its ability to incorporate AI into user interfaces and experiences. AI will deliver relevant information that could make this even more efficient, analysing details such as the customer’s location, the device they are using, their most recent interactions with the retailer, and even the time of the day they are most likely to make a purchase. Such detailed data is not only invaluable to securing a smooth checkout but can provide merchants with the ability to deepen their personalisation strategies with customers moving forward.
Digitalisation has meant that most organisations now use integrated payment systems to streamline the payment acceptance process and integrate automatic payment acceptance. This means that manual storage or the compilation of transaction data are no longer needed. Large language models will assist with collating and analysing payment data across multiple sources and locations, enabling organisations to find what they are looking for quickly and accurately.
It is easy to get caught up in the hyperbole around AI tools like ChatGPT and Bard, or equally to dismiss them as a passing trend, but the technology that enables them has application far and wide. Generative AI is complex, but the way that companies can use it is not. A simple prompt, or set of prompts, is all that it requires to deliver a wealth of insight. What payment company wouldn’t want to take advantage of it?