ING Bank builds bot to finally replace humans in clubby bond market

[LONDON] Dutch bank ING Groep NV is spinning off an artificial intelligence platform it developed to help its traders pick bonds to buy and sell.

The system, named Katana after a samurai sword, uses a machine-learning algorithm to scan millions of transactions and identify which bonds can be traded most efficiently. Katana is aiming to raise £3 million (S$5.3 million) to fund its growth as an independent company with half of that investment coming from ING, according to a statement seen by Bloomberg News.

The notoriously clubby world of corporate bond trading has proved resistant to the automation that's taken hold in other areas of finance such as equity trading. But financial firms and bond investors are starting to catch on, stepping up use of artificial intelligence to crunch reams of data.

"Every day an investor could have possibly 3 million trades to consider," said Santiago Braje, former head of credit trading at ING and now Chief Executive Officer of Katana. "You can't do that without some kind of machine assistance."

Katana started out as a tool for ING's emerging-market bond traders in 2017 and in its first year of use it helped the bank win 20 per cent more market-making transactions and reduce costs. It has since expanded the platform into a tool for asset managers to find and compare trade ideas. Dutch pension fund manager PGGM is among investors using the system and providing feedback to Katana, according to the statement.

It's the latest in a number of companies to come out of the bank's ING Labs unit, including Yolt, an app to manage money, and Cobase, a platform for companies to manage multiple bank accounts. The bank is investing in Katana via a €300 million (S$3504.8 million) venture capital arm focused on fintech.

BLOOMBERG