Consensus across the fixed income sphere remains that autonomous execution is off the table despite the technology’s potential to improve trading workflows.
Rebekah Riddell, head of trading at Canada Life Asset Management, and Sam Ratcliff, head of enterprise engineering at Man Group, agreed that human judgement remains essential as buy-side firms integrate artificial intelligence into their trading workflows.
AI offers opportunities to improve reporting, data preparation and analysis, however both speakers highlighted the need to distinguish the technology from traditional automation, particularly when it comes to execution.
“Having the trader and machine judgement constantly involved with AI is probably the only way I can see it beginning to work on the actual trading desk,” said Riddell, on the panel at the Fixed Income Leaders Summit.
“I think we as traders can use AI to for peripheral parts of our job on the reporting processes or cleaning the data or looking at data pre- and post-trade but for the actual execution I think you’re always going to need the human judgment.”
Ratcliff highlighted the distinction between AI and automation, arguing that AI is a tool firms can use to achieve specific objectives, rather than an objective in itself.
“AI is effectively a tool, automation is an objective,” he said. “So I think if you think about it in that sense there are many different tools which you can use to achieve automation. AI being one of those that can help you on that journey.”
The technology’s unpredictable nature, however, means firms must ensure its outputs are properly checked before they are used in trading workflows.
“One of the pieces that I think is key is the element of randomness that comes into it,” affirmed Ratcliff.
“I think a lot of what was being shown just now in terms of that human in the loop being there to check what comes out the other side of that tool I think is really important because of that randomness, that non-determinism that is part and parcel of that tool.”
Riddell also pointed to the inconsistency of AI-generated answers as a reason for caution when applying the technology to trading.
“We’ve all used AI and asked the same question twice, and gotten two very different answers. I think that’s why we need to be very cautious when we’re using it when we’re actually trading because automating is a defined set of rules so it makes you more confident.”
Controlled adoption
Explaining Canada Life AM’s approach specifically, Riddell highlighted its position as an asset manager within an insurance group, describing the firm as cautious by nature.
“Often we’re never the first to move in anything and are often described as massive oil tanker that slowly turns and takes look, we probably are on the more cautious side of things rather than rushing in.
“But, with that said, I don’t know many other insurance company asset management that have the enterprise AI system that we currently do have.”
On his side, Ratcliff described Man Group’s approach as controlled rather than conservative, with the firm seeking to make progress while remaining aware of the technology’s limitations.
“We know what we want to apply it to or are comfortable applying it to but we do want to make good progress. We are a tech-enabled asset manager […] We’re very proud of the technology that we have in house in all different areas of our firm. And this is no different.”
He added that firms must develop a sound understanding of AI rather than assume that any gap between early adopters and those moving more slowly will be easy to close.
“To achieve ‘technological sympathy’ and understanding you gain that by using it, but you also gain it by keeping up to date and just sort of taking an interest in it. I can see the point in terms of the gap will be easier to bridge. But I think you need to start from a point of knowledge.”