How AI (and GenAI) are transforming KYC in financial services
In 2024, European banks invested over 150 billion euros in artificial intelligence (source: Bpifrance). But behind these record numbers, a concrete question remains: how do AI (and GenAI) facilitate KYC for financial services? This is the topic explored in this episode of Parole d’Experts, with Alexandre, Head of Data at QuickSign.
A brief history of document processing
Understanding the evolution of document processing also means looking back at the major stages in the transmission of information:
- 3300 BC: the first clay tablets – writing is born
- 1450: Gutenberg invents the printing press – reproduction becomes automated
- 1929: the emergence of OCR – machines start to read
- 2020: generative AI – machines write, summarize, and contextualize
Today, a new milestone is reached: machines understand and make decisions in real time.
AI in financial services: what are the concrete use cases?
Among the main use cases for artificial intelligence in financial services, we find:
- Customer relations (chatbots, conversational agents, personalization)
- Fraud prevention (anomaly detection, behavioral scoring)
- Digital onboarding and automated identity verification (KYC)
It is this last use case that interests us here, as it perfectly illustrates the ongoing transformation.
AI + KYC: the ally for seamless journeys
Let’s take a very concrete example: extracting a name from a proof of address. Simple in appearance, this operation is actually very technical. There are three major approaches:
- OCR + coded rules: this is the basic method. We apply an OCR to extract all the text and explain to the machine that after “Mr.” or “Mrs.”, there will be a name.
- Supervised Machine Learning: learning from examples, more robust and more advanced. Here, we train an algorithm that uses both the image and its text to create the rules itself to extract the name.
- Generative AI (GenAI): contextual extraction via prompts, ultra-adaptive. We train a GenAI algorithm on a large volume of documents, and from there, we prompt it to extract the name.
GenAI is the most robust because it is capable of generalizing, but it poses a critical challenge: response time.
Industrializing AI: an often invisible challenge
“When you subscribe to a financial service, you expect an instantaneous response.”
— Alexandre, Head of Data at QuickSign
In the banking world, AI solutions must be scalable, secure, and instantaneous.
It is not the algorithm itself that makes the difference, but the infrastructure and its integration into seamless industrial processes.
The real challenge is achieving excellent computing performance even though these AI methods are very heavy. This is where QuickSign makes the difference, providing an industrialized, secure approach that complies with industry standards.
In conclusion: making intelligence invisible… yet effective
From clay tablets to generative AI, the history of documents is one of increasing automation.
But in financial services, this automation must remain invisible to the user while being perfectly mastered on the back-office side.
At QuickSign, we believe that artificial intelligence should not replace humans, but enable them to act better and faster – with complete confidence. And for that, we must constantly evolve our methods and infrastructures.
Written by Nicolas G.
