20 Oct 2023 3 min read

Trading: Can AI save time and cut costs?

By Ed Wicks , Sami Ragab

With a significant amount of time still spent on 'low-touch' trades, we're looking at whether machine learning could potentially deliver long-term benefits for clients.


The following is an extract from our latest CIO Outlook.

When anyone not directly involved in finance considers an investment trading floor, cinematic images of traders waving pieces of paper and running around furiously may come to mind.

Today’s reality is rather different: physical trading has been replaced by computer-based trading, which is faster, cheaper and more efficient. However, that’s not to say there are no further gains from automation to be had. In fact, we’re currently investigating the potential benefits of machine learning, a type of AI focused on the use of data and algorithms that imitates the way humans learn.

 Although most trades are now communicated to counterparties electronically, given that they are routed using financial information exchange protocols, we still divide trades broadly into two clear categories:

  • ‘Low-touch’ trades, a portion of which can be fully automated
  • ‘High-touch’ trades that require traders’ time, skills and attention

Four goals

It’s in the organising and executing of low-touch trades where we see the most potential for AI to deliver time and cost efficiencies. In this vein, our trading research team has established four clear, underlying goals, which are to see whether a machine-learning module (MLM) could:

  1. Accurately predict whether a trade will be high- or low-touch prior to trader involvement, to reduce the time spend on sorting trades
  2. Provide execution channel suggestions, adapting dynamically to market conditions
  3. Use model predictions to identify ‘hot spots’ where we expect a trade to be low-touch, but it is currently executed as high-touch (findings can be used to challenge and align strategy across the trading desk)
  4. Investigate the potential to automate with smart execution routing strategies

In short, the aim of our research is to deliver an MLM that can correctly classify the execution channel of each order, thereby allowing a greater proportion of traders’ time to focus on executing high-touch trades.

Early results from our initial model testing have been promising. We tested our model on a significant subset of 2022 trading desk data and found that it predicted the execution channel of trades with 93% accuracy (results shown in the graphic opposite). In particular, the model was trained to minimise prediction errors for high-touch trades (0.14%).


Our research into using AI for trading is still in its early stages, with a lot more testing both in London and across our trading desks globally required before a MLM could be fully deployed. But if the early signs are anything to go by, those days of time-intensive involvement in the sorting and execution of trades could soon be banished to history.

Selected trading systems providers have understood this trend and started to offer some integration between their platforms and trading desks’ in-house MLMs. This way, traders and clients have the potential to benefit from the best of both worlds: the convenience and reliability of the platforms, and the flexibility and innovation of their own AI capability.

The above is an extract from our latest CIO Outlook.

Ed Wicks

Head of Trading

Ed leads LGIM’s Global Trading team. The team is at the forefront of LGIM’s efforts to create better outcomes for clients through best execution. Prior to assuming this role, Ed oversaw the global equity trading team at LGIM. Ed joined LGIM in 2015 from BlackRock where he was responsible for designing and implementing trading strategies for the beta, LDI and transition groups. Prior to that he was a Delta 1 trader at J.P. Morgan, responsible for several index trading books and derivative market making activities. Ed graduated from Loughborough University and holds an MSc in Business Management as well as a BA (hons) in economics and politics.

Ed Wicks

Sami Ragab

Head of Trading Research

Sami’s primary responsibility is to develop LGIM’s trading research function to embrace a more quantitative and data-driven approach. Until 2018, he held the title of Quantitative Analyst providing fund and market analytics within the LDI team. Sami joined LGIM in 2013 from River and Mercantile Derivatives where he had a similar role. Prior to that, he worked for Mako Global, designing quantitative volatility arbitrage strategies, and at BGC Partners as an interest rate and inflation quantitative analyst. Sami graduated from Columbia University with an MA in mathematics of finance and holds an MSc in mathematics and computer science from ENSIIE.

Sami Ragab