The Impacts of AI on KYC Process

When we talk about KYC usually we think of banks, money, and payments, but KYC could have broader aspects and use cases beyond just “money”. KYC process has been adopted by governments and even applied for contractors, workers, and employees. But what exactly is KYC?

KYC, is all about truly knowing the identity of a customer and their potential risks. This isn’t an easy task for an organization to perform on every single customer, especially with the evolving regulations, cost pressure, legacy systems, and more regulatory scrutiny. The amount these organizations have been spending on regulations and compliance is not sustainable. So, definitely they need help to reduce cost, keep up with the expanding regulations, and avoid penalties — they need machine help, I would say.

The field of Artificial Intelligence (AI) can play a big role in the KYC process. It can streamline the regulation and compliance process, eliminate fraud to some degree, solve the human error, automate repeatable tasks, and help reduce time and cost. Here are some, but not all, points that would better illustrate the role of AI in the KYC process:

• Image processing: helps perform real-time scanning of proof documents, and improves the automated decision-making process. Here are some examples: 1- security features in official documents such as ID, driver’s license, or passport, … etc. 2- exploring font and pixels tampering in a document. 3- image manipulation/editing detection.
• Pattern recognition and cognitive engine: performing real-time discrepancy check for uploaded documents, and providing near-instantaneous output.
• Behaviour Analysis: collecting data from a chatbot, for example, and performing behaviour analysis on that data can help identify the customer potential risks, and take the right decision accordingly such as auto re-run a KYC on the customer, or alert the business with all the potential risks discovered so they can act accordingly.
• Link analysis: using AI analytical models to analyzing different types of objects and transactions and exploring associations and relationships between them would make the KYC process more efficient, and instantaneous.
• OCR: Using an intelligent data extraction engine and perform auto-filling of KYC documents provides an enhanced and quick customer onboarding.

During my web surfing, I’ve found a very interesting non-profit project (appducks api) that’s tackling some of the AI aspects I have mentioned above. They still have some work-in-progress that is very promising — I look forward to testing them soon. I have started testing their API using Postman, and I’ve got a very good result. So, I thought I would share with you here my test cases. It might help other developers to adopt some of these API endpoints in their next projects as it helped me with an application I’ve built recently.

• To test the api you need to install postman on your local machine.
• Go to appducks API and pick whichever endpoint you would like to use. It’s well documented.

Let’s start testing:

let’s take https://api.appducks.com/face/matcher endpoint. This endpoint takes two parameters: face1 and face2 and both are just images URLs.

so let’s pick two images: Image1 and Image2

here is the postman result:

That sounds cool! I found out that a distance that less than 0.5 seems to be the correct threshold to tell that the two images are for the same person. The appducks team, however, thinks the best threshold is 0.6

Yayy, simple and handy! I would love to see more of these APIs in the near future. I have used these API in one of the applications I sell on NativeBase Market and it worked superbly. Please check it out if you are interested!

Thanks for reading! ♡