Write-up by Rupert Walker, Managing Editor, GlobalTrading
Automation throughout the trading process is increasing rapidly, but there are roles for human agency, according to participants at a roundtable discussion in New York City.
Trade automation continues to be driven by regulatory requirements for best execution and stimulated by the rapid development of new technologies. It can help remove human bias, enhance surveillance, and explain trade execution in a systematic way and benchmark its performance.
Eventually artificial intelligence and machine learning will likely further reduce the role of human agency in the trade process, but now there is still demand for staff with quantitative skills, managerial expertise and even personal networks and market knowledge built on experience.
At its best, automation is about solving problems for clients, increasing trade execution efficiency and achieving scalability, agreed panellists at an Itiviti-sponsored roundtable discussion hosted by IEX at its office in the World Trade Center, Manhattan on 27 June.
One panellist recounted an anecdote from an electronic trading conference a few years ago when a speaker, with tongue in cheek (perhaps) predicted that soon dealing desks would be fully automated – and protected from meddlesome humans by a guard dog.
However, if automation goes wrong, problems can grow exponentially without human monitoring and override capability. In fact, financial firms have learned from experiences a decade ago and put in controls concurrent with technology installation since 2008, and the industry is now generally a safer place.
Third parties can be a better, more cost-efficient option for trade surveillance functions, such as identifying spoofing and front running, and making control adjustments. A lot of trading surveillance is already automated, but patterns are changing and bad behaviour is becoming more heinous and difficult to identify. In some markets, often the best that can be done is merely to flag a signal that a trader might be being spoofed or layered.
Automation is well-established in developed equities markets, and is increasingly deployed in the operational processes of passive funds to match indexes and reduce tracking errors.
Algorithms are sufficiently different and can execute diverse strategies, which ensures trading is not homogenous and one-directional – at least in normal market conditions. For instance, smaller fund management firms often tend to be very active with distinctive strategies implemented by their own homegrown algorithms.
Dealing desks, especially at large asset managers, can tailor their trading strategies to match the diverse styles, such as momentum or value-driven, of their portfolio managers and automate the processes. Regulation increasingly requires buy- and sell-side firms to explain and justify execution, especially outlier trades, in a systematic fashion. One consequence of greater compliance costs in illiquid transactions might be to force some firms out of business.
Banks, fund managers and vendors are applying similar technologies to other asset classes, including fixed income and foreign exchange, but the transfer is far from easy. Markets have their own idiosyncrasies, levels of liquidity and dealing practices; fixed income, in particular, suffers from sparse data.
Algorithms are only as good as their inputs, and sometimes a human is needed to identify and explain why an outcome is wrong, that is, to validate the data.
Indeed, accurate and relevant data inputs are essential, so a major challenge is to find clean data and institute consistent channels to access it. Too often data quality is compromised and inadequate, which should prompt vendor suppliers to gain an edge over competitors if they can provide reliable sources and feeds. Moreover, cybersecurity and lax controls are a perennial problem.
Many banks believe that they can differentiate themselves with clients by retaining a proprietary data management capability.
Besides, outsourcing data quality carries inherent risks because it implies extending complete trust to an external party and surrendering control over a vital part the automated system.
Machines and humans
On a practical level, it seems clear that humans at brokerages are still needed to manage risk, supervise systems, oversee the connection between clients and automated processes, and interpret regulation – lawyers will always be in demand somewhere.
Rarely do machines just speak to machines. Instead, automation simplifies, streamlines and codifies the trade order cycle, which helps sales-traders provide a better service for their clients.
In some cases, automated processes are introduced for traditional mainstream trades while high value transactions are performed manually. In addition, the development of more sophisticated algos and the introduction of artificial intelligence (AI) technology means that the dichotomy is actually reversing: humans handle the basic transactions and machines manage the exotic, high-value trades.
Although many banks have been working on aspects of AI, such as predictive analysis for several years, there is now more clarity about its wider potential and more structure about its employment. It is being used to identify pattern variations and implementation shortfalls, examine alternative scenarios and process more variables.
Machines learn from history rather than in real-time, but access to extensive contemporaneous data sources, including news and social media, mean that the history is fresh and relevant. They are more dynamic than previously, when they were rule-based and static.
Perhaps at some point humans can’t compete with machines. Automation is taking over time-consuming operational processes such as client on-boarding, and algorithms are already trading news as well as implementing portfolio strategies – which suggest that the investment manager’s role might become obsolete too.
Besides, if your job can be eliminated by a computer, then how long would you actually want to do that job? And new technology will continue to require complementary skill sets from people. Trading experience, networks and market savvy might diminish as job requirements, but quant expertise will be highly valued and, ultimately, essential.
We’d love to hear your feedback on this article. Please click here