The fraud section in my new book was great fun to write. I know you can’t and shouldn’t call fraud “fun” in any sense, but the case studies I researched were interesting, and often quite astonishing.

In one case I saw personally (which I couldn’t mention in any detail in the book) we discovered a fairly senior colleague, who everybody thought was a lovely, capable person, was actually involved in approving six-figure invoices from a fake supplier. The police thought this “firm” was probably linked to the “Russian mafia”, and we only found out about the fraud when the police discovered this gang was receiving large payments from my firm (and told us)!

Anyway, buying-related frauds can involve just internal staff, as in the case of fiddling your expenses or using the company charge card wrongly, or can be purely externally driven, as in the case of many “invoice misdirection” cases, or might involve both internal and external players. That third category is perhaps the most common and includes classic frauds such as overpayments to suppliers or biased supplier selection in return for bribes or inducements to the buyer.

But technology, artificial intelligence in particular, is helping to pick up some frauds through its ability to analyse huge amounts of data and spot trends, patterns, inconsistencies and oddities. I remember a presentation from two or three years back which talked about using AI to search through corporate payments or approvals. The idea was that you might find for instance a budget holder who always submitted an invoice for approval or payment on a Friday afternoon, when it might be scrutinised less carefully! Or someone who always makes purchases with a value of £9,999 if the cut-off for approval is £10K.

But more recently, I learnt of another interesting approach. In this case, the AI focus is on emails and documents that flow within the organisation and to external third parties. It has been developed by a firm called FACT360, which is led by Paddy Lawton, who founded, ran and then sold spend analytics software firm Spend360 to Coupa in 2017. I spoke to Lawton and fellow director Andy Slater to get a quick overview of what they’re up to.

Of their three core products, AI Forensics  is most relevant to buying-related fraud work. It analyses documents and emails and produces a network “map” of who is talking to who within an organisation and across organisational boundaries, including to suppliers, for instance. It generates insights from that communication flows as well as from the content of the messages themselves.

So for example, if you apply the analysis to Enron’s data, before that firm’s crash and disgrace, you can see that one particular person was at the centre of a major web of communication within the firm, even though he wasn’t apparently very senior. It turned out he controlled one of the technology “marketplaces” that enabled Enron to falsely claim to be making money on transactions. This analysis of what FACT360 calls “prestige” can tell you a lot about what is going on within an organisation, and who is really important or powerful. 

“And there are subtle changes in communication behaviour that occur and can be detected when actors plan and engage in covert activity” according to Slater.

One of the interesting corruption cases in my Bad Buying book tells the story of the Sainsbury’s supermarket potato buyer, who conspired over some years with a major supplier to pay over the odds for potatoes in return for bribes. Might Fact360 artificial intelligence have picked this up?  Probably, says Slater. It is likely that emails between the main players would have been more frequent than for other similar suppliers, or show different patterns in terms of timing or even use of language. There might have been more obvious clues in the content too.

Of course, knowing that your email trail could be used in his way might discourage fraudsters from using that medium, but there is always going to be some record of contact, unless the participants are using real secret service tactics! And the beauty of these emerging AI technologies such as FACT360 is that the user doesn’t need to know or define what they are looking for – the system will highlight where it finds potential “unknown unknowns”, as Donald Rumsfeld famously put it. 

We’re still at the early stages of understanding just how AI is gong to affect our lives, and it may be that some implications will not be positive for many of us. But using it to detect and deter fraud and corruption in our organisations – and reduce Bad Buying – must be one of the more positive aspects of this fascinating technology.