Artificial Intelligence can help in fight against fraud

Kickbacks and favours like flight tickets in return for business is just one example of the rot that has set in some industries.

By Shalini Verma

  • Follow us on
  • google-news
  • whatsapp
  • telegram

Published: Mon 17 Jun 2019, 8:00 PM

Last updated: Mon 17 Jun 2019, 10:52 PM

A group of monks were building a monastery in a village, with the help of a few masons. One of the monks spotted the head mason stealing some building material. She was very upset but chose not to confront him. Instead she reported the incident to the head of that monastery.
In his infinite wisdom, the head monk responded, "There is no need to be upset. It is natural for humans to steal when given a chance to get away with it. The reason why humans cheat is often human. This is why we should have processes in place so that they don't have the opportunity to steal."
Sounds like your everyday corporate governance mantra.
When I was advising a retail chain, the company executive revealed that the company had installed CCTV cameras above each cash counter, but the cameras were not working because of budgetary constraints. I was perplexed why they went through the trouble of installing cameras that pretty much did nothing. The company believed that the mere presence of cameras was a deterrent.
Deterrence is the reason why corporate processes exist. Yet, today they are failing because crooks are constantly finding creative ways to steal and lie. We give this some respectability by calling it "gaming the system" or "bypassing the system". Fraud in healthcare could look like doctors recommending a medicine, but what we missed was their collusion with pharma companies to misuse medical insurance. Familiar story? This is the brutal truth about human fraud in every industry.
In a large MNC, senior employees would routinely claim for travel expenses that included unofficial detours to cities that were not meant to be part of their official trip. The expense claims did not raise an alarm. On one occasion the employees had happily shared photos of their revelry on social media. Finance found unusual travel expenses claimed for the company's annual conference. They correlated the expenses with the employees' social media posts, and the culprits had little to say for themselves.
We hear of dodgy business practices occurring all around us. Kickbacks and favours like flight tickets in return for business is just one example of the rot that has set in some industries. Employees give money to HR recruiters to recommend their profile within the organisation. Corporate kickbacks are even harder to spot because they get institutionalised in euphemistic business language like promotional allowance, miscellaneous expenses, or legal fee. Imagine a scenario in which a company files a legal case against a middleman in a large deal, and the money paid in an out of court settlement is in reality bribe indirectly paid to the procurement officer of the end client.
A perceived opportunity, an imminent financial obligation and rationalisation of fraudulent behaviour - these are three reasons why people walk on the dark side. These motives do not necessarily show up in transactions. Hence, traditional rules-based systems are outwitted by fraudsters.
Can the shiny new AI technology play whistleblower? AI has executives ducking for cover because of the potential risks of job losses. Fortunately, there are other ways AI can cut costs. Human fraud is a powerful use case for AI innovation.
Because AI is capable of analysing a much larger dataset, it can find patterns that analysts may not be able to spot. It is possible for organisations such as retailers to train their AI system on regular shopping transactions of a customer. Thus, AI is able to focus on outlier detection or unusual behaviour to see if the credit card is being misused online. In peer group analysis, the AI system tries to see if your behaviour is different from that of your cohort with shared behaviour patterns. Let's say the cohort is employees with similar job functions in a country. Ever wondered why your doctor referred you for an expensive medical test? The focus of AI is less on transaction and more on behaviour to find correlations in a doctor's referrals and travel history. Throw in the insurance claim and medical notes into the mix.
AI also helps to cull out false positives that plague traditional fraud detection tools, which is such a waste of time for everyone. These are genuine transactions that are mistaken for fraud. If AI is trained in outlier detection over a large dataset, its accuracy level is higher.
The machine can learn on its own, in which case it may find new patterns that humans missed out. Or it can learn from specific cases you would train it on. Fraudsters will keep trying to outdo the system, so the AI system much be constantly trained.
Admittedly, it is hard to draw a line between fraud and acceptable relationship building that is specific to a culture, yet AI can at least play whistleblower for the Ombudsman to further verify. AI would not have human anxieties of getting witch-hunted. If AI's findings were misplaced, its algorithm would simply be retrained without any fuss. So, if you are looking for a strong application of AI in your organisation, fraud is a good place to start with AI.
Shalini Verma is the CEO of PIVOT technologies

More news from