Harnessing AI to Eliminate Inefficiencies and Enhance Decision-Making

While computers can perform some tasks as well as humans, and some specific ones even better and more accurately, AI is here to improve our lives in ways that we once thought impossible. One of those areas is efficiency.

Inefficiencies appear in various forms. A common cause is the endlessly slow, manual data checks in a sea of documents. There are numerous examples where manual verification is not only inefficient but impossible. For instance, fraudulent employers post fake job ads with very attractive conditions on major job listing sites, with the sole aim of collecting as much personal data as possible to resell it on the market. Large online job platforms with millions of advertisers could never hire enough people to manually verify each advertiser to determine whether they are engaging in fraud.

This is an area where AI can easily make a positive impact. AI can flag potential fraudsters by analyzing the activities of numerous accounts. Humans can then decide whether the flagged account should be suspended, but with this approach, they will only need to review a small number of suspicious accounts for prohibited activities.

Companies involved in credit card transactions already use AI to detect fraudulent activities. VISA uses a machine learning algorithm within its AI platform called Advanced Authorization to monitor suspicious activities in real time. This program saved them nearly $25 billion in 2019 alone.

Advanced Authorization considers several parameters:

• Whether the account has ever been used at the given merchant
• Type of transaction: in-app or online, contactless, chip, or magnetic stripe swipe
• Whether the transaction is occurring at an unusual time of day
• Whether the transaction amount is unusually high
• Whether the action aligns with other transactions on the same account

Their algorithm checks all these attributes in milliseconds and assesses the likelihood of a fraudulent transaction. VISA processes around 6 million transactions per hour. Imagine how many resources would be needed if these checks were done manually.

Inefficiencies can be found everywhere. At Bolton Royal Hospital, patients often had to wait more than six hours for a specialist to review their X-ray. This prompted Dr. Rizwan Malik, a physician interested in AI, to conduct a project aimed at speeding up the X-ray review process. He conducted a test where each chest X-ray was first reviewed by one of his students, then the AI gave its opinion on the same image, and finally, Dr. Malik himself reviewed the image. He wanted to determine how closely the AI's opinion matched his results. If the match was good, his intention was to use AI to check his students' work in the future.

However, before this test was completed, the COVID-19 pandemic struck. As chest X-rays became a highly effective method for identifying the most severe COVID-19 cases, the AI program was redirected to assist in quickly reading X-rays. This allowed for rapid determination of whether a patient had severe pneumonia, enabling them to be urgently referred for further treatment. In this case, AI significantly reduced waiting times for the most at-risk patients.

These examples show that AI is an excellent tool for automating time-consuming and repetitive processes that require human decision-making, as AI can be trained to make decisions almost as accurately in a fraction of a second.