The overwhelming narrative recently has been everything surrounding artificial intelligence (AI) and machine learning (ML). What do these technologies mean for society as a whole, and how will they be incorporated into daily life?
Large language models (LLMs) are a type of AI that can mimic human intelligence. They analyze vast amounts of data – learning patterns and connections between words and phrases. These LLMs are the algorithmic basis for generative AI chatbots like OpenAI’s ChatGPT and Google’s Bard. Generative AI is a form of machine learning that is able to produce text, video, images, and other types of content. A user inputs texts in the form of a prompt, then the LLM understands and generates human-like content in response.
The possibilities at the user level for this type of technology are endless – they can request writing prompts, ask difficult research questions, input math equations, even receive coding help. The applications for AI at the enterprise level also continue to explode. Numerous Couchbase customers, across a wide range of industries, are planning to utilize AI in their businesses to create a better experience for their customers.
Artificial Intelligence in the financial services industry
One area where AI is already playing a pivotal role is within the financial services industry to combat fraud. Fraud is arguably one of the biggest ongoing challenges for financial companies due to the growing opportunities to exploit the less mature technical infrastructure of many fintech startups. Already costing the global economy £3.89 trillion (around $4.5 trillion in USD), the impacts of financial fraud are only set to worsen unless fintechs implement the right technologies.
Historically, detecting and preventing fraud can be an expensive, labor-intensive process for online fintech startups since they often lack the extensive fraud departments and call centers of traditional banks.
Revolut uses AI and ML to strengthen fraud detection
Revolut needed a fully automated system that could identify fraudulent transactions, notify customers, and allow or block payments without human intervention. To do this, the company developed Sherlock – a machine learning-based card fraud prevention system – to counter the growing threat of financial fraud.
Sherlock continuously and autonomously monitors transactions in less than 50ms for over 12 million customers. If it deems a transaction suspicious, it blocks the purchase, freezes the customer’s card, and sends a push notification prompting the customer to confirm whether the transaction was fraudulent or not. If the customer responds that it was legitimate, the card is unblocked, and they can simply repeat the purchase. However, if the customer doesn’t recognize the transaction, the card gets terminated and they can order a free card replacement.
Sherlock is built on Couchbase’s NoSQL database, where user and merchant profiles are stored and ready to be retrieved at the moment of evaluating whether a transaction is fraudulent or not. With data on users and merchants changing rapidly, Revolut needed a database that had the speed and agility to react quickly, with the scalability to handle millions of documents.
The results of such innovation have been incredible: thanks to Sherlock’s performance, more than $3 million per year of customers’ money is saved through preventing fraudulent transactions, with just 1 cent out of every $100 lost due to fraud – compared to an industry average of around 7-8 cents. Sherlock flags and prevents over 96+% of fraudulent transactions and continues to refine its machine learning algorithms even further.
FICO and Wells Fargo protect customers from fraudulent transactions
FICO and Wells Fargo are Couchbase customers with similar stories – utilizing AI to help protect customers from fraudulent charges. FICO’s Falcon Fraud Manager is widely considered to be the #1 fraud detection platform in the world and scores 65% of the world’s credit/debit cards.
Falcon Fraud Manager monitors transactions end-to-end to detect and prevent fraud in credit cards, debit cards, prepaid cards, commercial cards, digital payments (including real-time, P2P applications like Zelle, Venmo, FedNow, CashApp, etc.), account-to-account, and wire transfers.
Downtime means fraud and lost revenue for the company, so when FICO was chosen to provide credit checks, fraud screening, and targeted offers for new telecommunications customers both in-store and online, it needed a NoSQL database that could deliver high availability alongside high transaction volume. Couchbase was chosen over Cassandra and MongoDBâ„¢ for speed, scalability, availability, and persistence to support large XML objects.
Wells Fargo utilizes FICO’s Falcon Fraud Manager, along with Couchbase, to support its fraud monitoring infrastructure. Wells Fargo applied machine learning analytics to internal and third-party data to identify and adapt to sophisticated fraud attacks in real-time. Now, 100% of transactions are processed in real-time for fraud – totaling 50+ million transactions per day – at less than 10 ms per operation.
This is just the beginning for these types of use cases. Couchbase customers continue to be on the forefront of innovation, incorporating AI into their businesses to create a better experience for their customers.