Source: IFLR
One specific segment of financial services which could benefit most from AI is the fixed income space, which has historically been considered too complex for automated technology. Around 60% of European government bonds and 50% of US treasuries are traded electronically, and new regulation could push this further.
Mifid II requires firms to report prices and volumes of completed bond transactions, a highly resource-intensive process unless some kind of automation is implemented. The best execution requirements for fixed income also encourages algorithmic trading to provide an audit trial.
“The fixed income asset class is seeing significant efforts to automate the execution process with some dealers fully automating the execution of odd lots,” said Munder Shuhumi, managing director of Algoritmi Investment Solutions. “This will make the asset class more accessible to robo-advisors.”
Following Mifid II, designed to increase transparency in financial services in the aftermath of the financial crisis, buy-side companies are being forced into purchasing their own research. Before the directive came into
force, buy-side firms purchased research from investment banks and other third parties, without needing to stipulate where the data came from, and would then pass on the cost to investors. Now forbidden to do so because of so-called research unbundling, the rule change could encourage banks to utilise AI to make it easier when data-gathering.
KEY TAKEAWAYS:
- Mifid II could encourage financial services companies to increase their adoption of AI, particularly in the fixed income space;
- The fixed income space has historically been reluctant to use automation, but this change could push other segments to use machine learning tools;
- The FCA is trying to push the use of AI in financial services and is having success at the moment.
Technologically cautious
Yet banks are notoriously slow and conservative in this area, a fact which has been exploited in recent years by challenger banks and neo-banks. Shuhumi believes deep learning tools provides greater transparency and confidence in decision making, but it does not have a big enough track record. Only then will mass adoption be made.
“Robo-advisors are currently shown to be underperforming human advisors, but once the transparency is improved and a successful track record has been established, investors will probably increase their allocation to robo-advisors, which could increase dependency on pure AI advisors,” he said.
Robo-advisors, those providing advice with limited human assistance, are gaining in popularity. NatWest, Santander and HSBC are launching one of some kind. But without a track record, banks are unlikely to be fully committing, even if it costs less than a human advisor.
In a study by Boring Money into robo-advisor portfolios last year, based on a £5,000 ($6,464 approximately) investment only three beat the FTSE100.
But with time, this performance could improve. Cultivating machine learning algorithms makes it significantly easier for novices to access AI and increases the number of people working in the field, making it far more
accessible. And if a segment of the financial services sector, once so reticent to adopt automation, begins to adopt it then it could well be a signal for other segments to adopt deep learning tools themselves.
“We understand regulators need to make the case for transparency, but this needs to be incentive-based,” Shuhumi said.
A surge in regulation this year is likely to see renewed investment in artificial intelligence. Much of the technology utilised so far has been on basic AI, but the real benefits come from deep learning, where machines
effectively teach themselves entirely unsupervised through advanced algorithms. This could be particularly advantageous for trading, where circumstances change so quickly that it is very difficult for a human to keep up.
Many possible uses
Tara Waters, partner at Ashurst, believes AI has been under-utilised across the industry so far, but with regulation of this kind there could soon be a change in focus.
“Much of the focus so far has been on anti-money laundering and know-your-customer processes to reduce manual and redundant inputs; we are also seeing backend review and documentation being impacted,” she said. “The next stage is when firms can use machine learning to assess their compliance across a multitude of laws and regulations, including through automated review of materials.”
The UK Financial Conduct Authority (FCA) together with 11 other global financial regulators created the Global Financial Innovation Network last month, with the aim of providing a more efficient way for firms to interact with regulators. It is also said to create a new framework for cooperation between financial services regulators.
In July, Nick Cook, head of data and information operations at the FCA said they were considering the possibility of making its handbook machine-readable and machine-executable, so machines can interpret the rules themselves.
“The FCA has embraced new technology and is very conscious of not stifling innovation,” Waters said. “This is the spirit needed for future regulation, because we need to encourage new technologies and new entrants in the market, all with the goal of increasing innovation.”
