Today, Artificial Intelligence (AI) is extensively used for complex problem solving in real time business applications and is recognized as a value add in the financial services industry. Specific to wholesale banking, we are aware of the gaps that exist. In some cases, payments made by buyers to corporate customers of the bank, lack customer and invoice information. Thus, leading to an inability to reconcile customer accounts and generate month end reports. Most of the popular ERPs can reconcile if the customer is identified against the payment made. Since that is not possible, the process involves manually identifying the customer and applying the payment against the invoices raised by him/her. It is something akin to finding a needle in a haystack.

For example, we have all seen movies that show advanced biometric devices that overlay a fingerprint impression over a few thousand samples for self identification in order to provide or discard access. In reality, it is impossible to prove an identity through this method. In a country like India, imagine 1.2 billion people being asked to overlay fingerprint impressions from the database of a billion+ samples to identify someone. That would be a hugely complex and humungous task. The process becomes much simpler when the individual is identified in the system and you just have to match the current fingerprint sample with what you already have in your database.

In the case of wholesale banking, the problem is some what similar. When a payment is made by a buyer, the following scenarios are possible – customer/invoice details or both are missing. If we know the invoice number or customer identification then it is a straight-forward case like the example we discussed above, where we map the current fingerprint sample with what we have in the database and validate it. If both customer and invoice details are missing, the problem becomes very complex. It’s even more difficult when the details (customer and invoice) captured are spelt incorrectly. Research might help create matching engines that can map receivables to invoices.

Does all this mean one can replicate Warren Buffett’s business acumen? It might be possible in the distant future. Kathleen McKeown (Professor of Computer Science at Columbia University) is imagining creating an “electronic” version of Warren Buffett. Good luck, Kathleen!