Of the $30bn invested globally in fintech in 2017, just 6% went to primary and secondary capital markets. While payment innovations based on distributed ledger technology (DLT) and open application programming interfaces (APIs) power ahead and retail banks experiment with biometric technologies such as facial recognition, corporate and investment banks (CIBs) are often portrayed as a slow adopter of fintech.
CIBs’ back-office tasks have been automated for years to cut costs, reduce human error and improve efficiencies. But many in the industry acknowledge that fintech has had less of an impact on their revenue-generating side.
More than meets the eye
CIBs’ sheer complexity means a relatively small proportion of fintech firms and financiers dedicate themselves to the area. “Most investors have a grounding in tech rather than finance, and while business-to-consumer areas such as lending and payments are easier to understand, very few understand the investment banking world,” says Huy Nguyen Trieu, a former Citi managing director who co-founded educational platform the Centre for Finance, Technology and Entrepreneurship.
This is exacerbated by some investment bankers, who believe their relationship-driven businesses are impervious to tech disruption, and by the fact that regulatory technology, or reg tech, is being prioritized to satisfy market overseers. “A higher proportion of [CIBs’] tech budgets are having to be invested in compliance and regulation,” says Warren Breakstone, managing director at S&P Global Market Intelligence. “What has suffered as a result is the amount being allocated to revenue-generating innovation and fintech.”
Oliver Harris, head of JPMorgan’s wholesale banking fintech programme In-Residence, agrees that the fintech story began in retail, before moving into asset management and now capital markets. But he believes the collaboration with CIBs is more buoyant than is widely perceived. “The reason investment banks may appear to be doing less is perhaps because the ticket size of the investments tends to be smaller as the innovation is just starting to emerge. And helping fintechs test their ideas at an enterprise level can take time,” he says.
Given that industry revenues have been flat for a decade, CIB front offices need a helping hand from fintech now more than ever. The solutions being tested and adopted are often niche, and they belie the buzzword categories such as artificial intelligence (AI) and machine learning, but they show great potential. The Banker takes a look inside some of the most forward-thinking CIBs, to see what they have been up to.
Markets divisions have taken the biggest strides. The move to electronic platforms over the past two decades – combined with trading’s reliance on processes, mathematics and timing – makes them ripe for automation and advanced analytics. The benchmark has been set by JPMorgan in equities, which is unsurprising given that it was among the first asset classes to digitize in the late 1990s. The bank has embedded machine learning into its equities algorithms that place limit orders (instructions to buy or sell stocks at a specified price or better). Based on historical trade data, the programme known as LOXM runs millions of market scenarios in real time to place trades at the best time and price for the client. It is one of the first examples of AI being used to execute trades.
In the bonds universe, large corporate trades are still made over the phone, but small trades have moved to electronic platforms (e-platforms). The result is that some 90% of tickets now trade electronically, a large proportion of which are neglected by sell-side traders who must prioritize the bigger voice orders.
In response, Julian Pomfret-Pudelsky, a fixed-income algorithmic trader at Credit Suisse, created an algorithm called CSLiveEx, which autonomously handles inquiries of less than a notional $1m. “It decides what price to respond with and subsequently, if it wins the trade, what to do with the risk,” he says. “It basically emulates what an experienced institutional trader would do, except faster, more efficiently, and with better accuracy.” CSLiveEx enables the bank to offer a price on 10 times more qualifying trades than in the past, while the number of trades it wins has trebled.
As the number of e-platforms has proliferated, so too have the number of APIs that CIBs must create to plug into different venues. A London-based intermediary called TransFICC has solved this by allowing CIBs to connect to a standard interface which in turn connects them to multiple e-venues.
Fixed income’s fragmentation into voice and e-trading makes it harder for sales teams to keep track of client activity across the entire market. JPMorgan has addressed this by partnering with data analytics firm Mosaic Smart Data, which integrates the bank’s fixed-income voice-data and e-data into a single platform.
“With the increasing electronification of certain macro products, the ability of a sales person to see their entire client flow in a single place is fantastic,” says Mr. Harris. “They shouldn’t have to log into multiple systems to understand how much they are doing via e-platforms and how much by voice.” Mosaic can also recommend suitable trades based on real-time analysis of clients’ activity. It is akin to Amazon’s powerful recommendation engine, which McKinsey estimates generate 35% of its sales.
Pricing is among the front-office trading activities that stand to benefit most from fintech. A good example is ING’s AI-powered tool Katana, which uses advanced analytics on historical and real-time trade data to help bond traders decide when to transact and at what price. When tested on the bank’s emerging market desk in London, Katana led to faster pricing decisions 90% of the time and a better price four times more frequently.
Underpinning many trading advancements is faster computing power, an area pioneered by Maxeler. It builds so-called supercomputers which use dataflow engines, in which each component works simultaneously, rather than traditional control-flow engines which instruct the components sequentially. “Compared to a grid solution or server farm it means [traders] can run risk calculations almost instantaneously to see, for instance, what their exposure is in their large books when the market moves,” says chairman Geoff Smailes. JPMorgan’s fixed-income team started using a supercomputer in 2011, which has reduced calculations on its structured product book from about eight hours to two minutes.
CIBs are investing heavily in alternative data, whereby they extract value from non-market sources. By hiring data scientists and partnering with firms that analyze everything from satellite images to shipping traffic, job openings to the furthest corners of the internet, traders can now spot trends and correlations that help them make better informed and faster decisions. The value of alternative datasets has become greater as reforms make markets more transparent.
“Pretty much everyone now has access to [market] data so the chances of generating any alpha are close to zero,” says Axel Pierron, co-founder of consultancy Opimas.
A popular application is sentiment indices, whereby firms scrape social media and blogs to create a gauge of public opinion on different companies. Equity traders use these to help forecast if certain stocks will rise or fall. New York-based Selerity has pioneered another application, which uses natural language processing (NLP) to search and tag social media and newsfeeds to generate highly tailored search results for those looking to trade.
Citi has onboarded Selerity to process external data, while the bank applies NLP to its internal data. It then merges them into a single set of search results which it provides on its digital execution platform Velocity, allowing informed decisions to be made more rapidly.
“It has the ability to extract actionable knowledge or insights from data by filtering, consolidating and condensing information on a particular topic,” says Stuart Riley, Citi Markets and Securities Services’ global head of operations and technology. “That may not sound difficult, but tailoring it so that the output ends up specific and reliable enough for a salesperson, trader or client to make informed decisions off the back of it can get very complex.”
Alternative data is not, however, just about assessing securities. Sigma Ratings, which participated in Barclays’ US accelerator, applies AI techniques to different types of data to help traders assess their counterparty’s non-credit risks, such as financial crime risk. “Our and regulators’ expectations for a robust and comprehensive set of indicators regarding counterparty risk will continue to grow,” says Nej D’Jelal, Barclays’ co-head of electronic equities for Europe and the Middle East. “When there is a company whose day job it is to collect, normalize, store and make actionable these various data sets, it is important to collaborate to enhance traditional approaches.”
Capital markets and loan origination trail trading in adopting fintech. Bankers say this is because their work involves less process and more advice. But some admit they received a wake-up call in early 2017 when Ipreo, which hosts the bookbuilding interface between CIBs, electronified the bond sales process by launching Investor Access, an online portal via which investors and syndicates now communicate. Developments since then have established syndicate desks as the fintech leaders in the new issue market.
Seven banks are working with tech firm Finastra to create an online marketplace for syndicated lending deals using software firm R3’s blockchain platform. It will create more transparency and efficiency between agent and lender banks, which today communicate via phone and e-mail. Meanwhile, ING has partnered with Italian start-up Axyon AI, whose SynFinance platform uses deep learning techniques to help bankers identify investors most likely to participate in syndicated and leveraged loans.
While deep learning is behind some of the biggest AI advancements in retail banking, Axyon CEO Daniele Grassi believes it also has great potential for CIBs. “It can lead to a better understanding of the reasoning and behaviour of market players, and to more accurate predictions regarding market trends. In the primary market it can help assess deals’ values and risks, and structure them in a more data-driven way,” he says.
As with trading, primary market pricing is ripe for fintech enhancement. Today it is a time-consuming, human-driven process whereby bankers integrate primary and secondary market data regarding the issuer and similar credits. With debt markets approaching the end of their cycle, new data analytic technologies are needed now as never before. “The more data that goes into forming your advice, the more you reduce execution risk, which in the end is the most important thing,” says Armin Peter, UBS global head of syndicate. “Having spent the past decade in a bull market we’ve had a pretty easy life in terms of transacting successfully, but when markets turn, the question of how technology can help us becomes particularly important.”
In terms of structuring, Crédit Agricole has onboarded France’s QuantCube Technology, which uses AI and alternative data to overcome the three-month time lag in official macroeconomic figures. Its creation of inflation figures is a good example. “In reality, [its] variation can be explained by four factors: fruit and vegetables, energy, travel packages and clothes,” says QuantCube CEO Thanh-Long Huynh. “Having access to data that predicts the variation of inflation in real time, based on those four components, is useful for bankers working on inflation-linked bonds.”
Meanwhile, JPMorgan is at fintech’s cutting edge in helping bankers generate leads. In 2016 it launched its Emerging Opportunities Engine, which uses machine learning to analyze clients’ transaction history, market conditions and the way different industries have evolved to identify which clients are well positioned to raise equity or debt.
Advisory not immune
It is understood that other CIBs are planning, or already building, similar recommendation engines for capital markets as well as mergers and acquisitions (M&A). The latter has proved particularly resistant to fintech, with bankers often arguing that their primary assets – client relationships fostered over many years – cannot be replicated by technology.
ING wholesale banking’s head of innovation, Ivar Wiersma, agrees up to a point. “The delivery of advice and the trust and personal relationships that drive complex decisions in M&A scenarios make it likely that the human factor will remain very important here,” he says. “For distress and restructuring situations I see a similar scenario.” But Mr. Wiersma notes that a lot of M&A analysis (plus pitchbooks and presentations) can be automated. AI has great potential here, but it will materialize later than in other parts of CIB.
While recommendation engines suggest new opportunities based on external data, CIBs are also looking to integrate – and then leverage – internal data to enhance sales teams’ client knowledge and ability to suggest suitable services. “Everyone talks about having a 360-degree view of a client, but the reality is they are far from that,” says Mr. Pierron. “A lot of the solutions being developed in this space don’t require big infrastructure changes, which means the implementation risk is minimal, but these projects are taking a long time.” JPMorgan’s Mr. Harris notes that the In-Residence programme is partnering with JPMorgan’s senior leadership to think through fintech’s collaboration with M&A.
Citi’s Mr. Riley holds the consensus view that in the long run, data analytics in its various permutations will have a bigger impact on markets than any other type of fintech. Many traders and investment bankers are skeptical regarding DLT (some believe it solves problems that other technologies can tackle) but Mr. Wiersma believes that it, along with AI, will have an enormous impact on CIBs in the long term. “People tend to overestimate the impact of new technologies in the short run but underestimate the impact in the long run. For DLT I think that is definitely the case,” he says.
The perennial question, however, is to what extent fintech will oust the CIB front office. Mr. Nguyen Trieu says fears that AI could replace traders have evaporated because very few CIBs have been able to create alpha from algorithmic trading using alternative data. Banks claim they are not using technology to downsize their front office, but rather to enhance what traders and bankers can deliver and allow them to focus on value-added work.
Yet Alex Manson, global head of Standard Chartered’s new fintech investment and innovation unit SC Ventures, takes a more realistic view: CIBs cannot escape the universal truth that technology makes some jobs obsolete. “It applies to a bank’s front office in the same way it has applied to society in general for years, going right back to the Industrial Revolution,” he says. “As in the past, the best way to respond to a rising tide of technology that has the ability to replace jobs is through education. It’s not that we no longer need humans, we need humans with a different skillset.”