Johannesburg, 21 Feb 2019
The recent Bosasa allegations raised at the Zondo Commission suggest vast amounts of fraud and corruption have taken place in the many government contracts undertaken by this organisation. Knowing this, it is certainly worth asking whether the implementation of the right type of technology might have been able to more quickly uncover such activities.
According to Mint Technologies' Head of AI, Peter Reid, artificial intelligence (AI) and machine learning offer the potential to help uncover potentially fraudulent activities conducted between two business parties.
"With corruption, it's generally a case of the criminals constantly trying to find ways to get around the controls and structures put in place to prevent criminal activity. As a result, companies put in place stricter procurement processes, until such time as the criminals find a way around those, after which they need to refine and adjust their processes again, in a never-ending 'arms race'," he explains.
"However, with something like AI, it becomes possible to consider extremely large data sets in order to pick up patterns that are unusual. The larger the data set, the more complex the pattern that can be discovered. For example, if a particular suspect has purchased an expensive new car, it is simple enough for AI to look at the information available to determine whether there has been a corresponding dip in their lifestyle and spending, compared to millions of other people that legitimately bought expensive new cars. Not only is this something that would be impossible for a human to determine in a reasonable length of time, but with technology like AI, it is possible to add many other parameters to the equation."
Reid points out that all corrupt activities, including those alleged to have occurred in the public sector, require two parties to be involved: the one offering the bribe and the one accepting it. He adds that since there is an enormous amount of company data available, it might actually be easier to discover those handing out the bribes instead of those accepting them.
"Of course, it's also worth noting that training AI is a complex task, fraught with difficulties, not the least of which is the importance of having clean data to work from. Remember that if you input biased information, the AI will inevitably produce biased answers. Therefore, it's vital to train the system carefully and correctly, before implementing it."
"Moreover, for your AI to work effectively, you need to ensure the right technological foundation is in place. Firstly, this means vast quantities of clean data, both structured and unstructured, not to mention an effective cloud strategy, in order to store and process your data inexpensively. Generally, when it comes to data, my advice would be: when in doubt, store it. Even if you cannot see the benefit such data can offer, AI often picks up the strongest patterns from the least likely sources."
Of course, he continues, the most important requirement is for the business implementing the AI to clearly define the answer it is looking for. This could be a query relating to something like fraud or corruption, or it could be something else of strategic value to the business, which was previously unanswerable. For example, what is that one thing that could significantly and positively change your business if you knew the answer to it?
"Personally, I don't think we're too far away from a future where machines are able to learn how to perform complex jobs well enough that the entire procurement and supply chain environment can be left in their hands. Obviously, the fewer people involved in the chain, the fewer potential points where criminal activity might occur.
"Therefore, as machines become more advanced at analysis, so businesses will be able to reduce the human factor in their supply and procurement chains and free them up to work elsewhere in the organisation. Moreover, looking a little further into a future where we are already anticipating things like self-driving cars, it seems we'll have intelligent machines undertaking a range of jobs that currently require a human. And taken to its logical extreme, this points to a future where we may well leave the complex administrative issues related to running an entire country to a machine, instead of a human. This will have the added benefit of enabling humans to move to higher-value positions," he concludes.
Share