As AI and automation begins to ramp up across the industry, experts at the Fixed Income Leaders Summit debated where the line is drawn between the machine and the human, and whether we should truly begin to automate human judgement.
As explained on the panel, automation provides many opportunities in boosting execution quality, with one expert asserting: “The two key benefits for us of automation are one, efficiency, especially when you’re running very lean trading desks. Two, the capacity to free up the trader’s time to focus on challenging trades.”
Despite this, limits remain, and human intervention remains a key part of fixed income automation.
Read more – Where does the growth of AI leave the role of the human and the trader?
In particular, human intervention in automation is largely dependent on product type in fixed income markets – with credit markets often tougher to automate due to dealer selection complexities and execution criteria nuances.
Similarly, portfolio management were highlighted as an area which allows automation to move up the chain from portfolio managers to traders to support the trading desks, although this again can surface complications.
As one panellist explained: “You can have a lot of competing PMs who are eager to push the limits, eager to trade a lot and do more. But then we also need to have a relationship with the dealers as well. So it’s complicated.
“PMs are not used to trading in the market, and they don’t have the experience of how to work with the dealers. So we try to automate the judgment of PMs because they’re not really interested in going deep into how to execute.”
An ‘assessment management’ business
Building on this, discussions also turned to ensuring automation is used in the right capacity, and human judgement remains critical when automation struggles, particularly in less liquid or complex trades.
As one expert explained, this comes down to ‘assessment management,’ – they added: “We embrace automation and innovation, however, automation is just one useful tool amongst an array that’s available to the trading desk to use. We automate what we think makes sense depending on the market structure and liquidity of each product.”
Similar sentiment was echoed when it comes to the real-time trading process, with one panellist emphasising that having an audited model which is tightly controlled is key in these scenarios.
“[…] this way we understand how it works, we can predict how it will work going forward an if it does something we don’t like, we can make an improvement to it, versus throwing it in an AI black box. If your self-driving car crashes, how do you troubleshoot that?”
Moreover, human intervention also remains important in fixed income to ensure relationships between the buy- and sell-side are retained.
Specifically, maintaining the ‘human touch’ element when it comes to AI and automation provides traders with the flexibility and experiences to observe what banks are pricing, and how the market is moving.
Explaining this, one panellist added: “Fixed income trading is a relationship business. We do want traders to have good mitigation skills and great relationships with the sell-side, and be plugged in to be able to determine the trading strategy.”