Jumio, the AI-powered trusted identity as a service provider, announced gains in the speed and accuracy of its verification services, as well as a more intuitive user experience.

These gains come after a two year period during which Jumio invested heavily automation enabled through a variety of machine learning, artificial intelligence, and optical character recognition (OCR).

As a result of its investment in supervised machine learning models, Jumio was also able to recently launch Jumio Go, its new identity verification solution powered exclusively by AI. The investment has also improved Jumio’s current suite of identity verification and authentication services.

“We’re seeing across-the-board improvements in our ability to automate virtually every phase of the identity verification process, making our core solutions even faster, easier and more accurate for our customers and their end users,” said Labhesh Patel, Jumio CTO and chief scientist.

Jumio is highlighting a number of improvements that have come about as a result of its focus over the past two years. Its AI, machine learning, and OCR improvements has increased identity verification response times by 33%, and additionally almost 90% of the verification process is now fully automated. This means that verification experts are only needed in cases where the image quality of a document is poor and requires human verification to make a definitive identification judgment.

The system also gives users the ability to course correct — where they are given specific reasons as to why their ID or selfie was rejected so they can make adjustments and continue with their registration — which has increased conversion rates by 15%. Additionally, users have the ability to complete registration across devices now, so a process started with a selfie on their smartphone can be securely completed on a computer with the use of an SMS message or QR code.

Jumio has also integrated with FaceTec ZoOm, the first biometric tech to achieve Level-1 and Level-2 certification through iBeta Presentation Attack Detection (PAD) testing, for liveness detection that can detect deepfakes and online spoofing attempts.