Generative artificial intelligence (GENAI) and agent models revolutionize financial services, transform user experience, fraud detection, innovation of payments And compliance process, according to John KainFinancial Services Development Manager at Amazon web services (AWS).
“The AI has been an integral part of how industry has modernized in the past decade. Really is in all parts of the value chain of financial services, “said the former JPMorgan leader in an interview with Pymnts.” What has radically changed in the past two years is the impact of generative AI on all these processes. »»
Kain sees three essential trends between banks, capital markets, insurance And Payments:
- End -to -end user experience: AI allows transparent integration, product recommendations, maintenance And Rear treatment. He is also powered hyper-personalized experiences, empowering companies to adapt offers, advice And communications in real time Data -based and user behavior.
- Data modernization: Inherited systems are upgraded to unlock basic data, allowing more personalized services and better analyzes.
- Integrated financial services: Companies incorporate financial tools into wider value chains, meeting users in any channel, whether in electronic commerce, social platforms Or mobile applications.
In the payment space, Kain underlined the global rise in real -time payment systems, from the UPI of India (unified payment interface) to the Brazil Pix, which offer Instant establishments and low fresh. In the United States, banks are now leading, according to a Pymnts intelligence report in December.
“These changes transform customer expectations,” said Kain, adding that innovations like buying now, paying later and Stablecoin rails create new challenges for infrastructure and fraud prevention.
Automatic learning is deployed to improve fraud detection – especially for instant payments, where there is little room for error. AWS also notes the adoption of techniques such as the formation of a model distributed for fraud and white room environments to share fraud data between institutions.
Learn more:: More than half of American companies use real -time payments
The agentic AI is gaining ground
Kain said IA agents – AI systems that are looking for, interact with other agents And Complete tasks for the user – are “certainly the way the industry takes place”. Current use cases is agents who access internal information to serve customers more efficiently.
Kain cited the case of WithdrawA money transfer company portion Transactions in 18 languages. For 95% of funds, How’s it going directly without problem.
But For the remaining 5%There could be delays Due to the identification of customers and other wrinkles, who erodes confidenceaccording to rhyme. The company uses the GENAI to help resolve cases by things like Find the right internal information to help the customer.
Other use cases include:
- Call center automation: Real -time transcription, feeling detection and routing to digital Canals.
- Compliance: Accelerate anti-flash surveys (AML) on alerts, in a case of 80% to 90% using the LLMS.
- Personalization: Banks using AI can adapt the marketing not only of products for the customer, but also optimal means of paying.
When it comes to precision Answers and hallucinations, which are essential for financial services institutions, Kain said that the models have improved on this front according to generation (RAG) (RAG) and other recovery techniques and other as well as to have a knowledge base of confidence information. In addition, the AWS Genai platform, the rocky substratum, has railings that use Genai to detect hallucinations with a success rate of 75%.
However, “there is still a heavy human aspect” with regard to important financial decisionsKain said.
Recognizing concerns about the cost of the Genai, Kain detailed AWS efforts to reduce spending through personalized fleas (Infentia and trainium), the model distillation and the tools for comparing flexible models. Smaller and distilled models adapted to specific tasks reduce costs while preserving high performance.
Kain also said that customers can Choose models such as Nova, the own AWS family of large languages models (LLM), which is 75% cheaper than comparable models of competitors when it was unveiled last December.
Deepseek R1 is also available on AWS, from the Chinese AI startup which bumded Silicon Valley and Wall Street for its economical model. Kain said companies can use Deepseek without worrying about sending their data abroad. As it is hosted on AWS, the data remains with the customer.
“We allow customers to test different models and prices so that they can optimize performance and costs,” he said.
For example, Nasdaq used An LLM for AML tasks but found the cost too high. It turns out that they unnecessarily used a very large model at the same time for the summary and to write the final report that goes to a regulator or their internal audit teams.
Nasdaq discovered that by using a much smaller language model to summarize the facts but to use The large model to write the report, they can always maintain the same quality and reduce the cost, said Kain. “By mixing the two, they could have been Find something that had an economic meaning. »»
More customers will find their own optimal approach. “I think you just will continue to see much more flexibility,” said Kain.