There is so much excitement around the potential for artificial intelligence and machine learning in finance to transform the way credit is priced, capital is allocated and products are sold that it requires a clear head to separate the conceptual debate over long-term possibilities from the reality of what is actually going on.
In July, Christopher Woolard, executive director of strategy and competition at the UK’s Financial Conduct Authority (FCA), spoke at The Alan Turing Institute's AI ethics in the financial sector conference.
“Superficial debate around AI and machine learning often descends into talk of robot armies and a dystopian decline in human agency,” says Woolard. “But are we really living through a crisis of algorithmic control? The answer, at least when it comes to the use of AI in financial services, is: not yet.”
The FCA recently carried out a joint survey with the Bank of England to assess the current state of play. “What we found was that the use of AI in the firms we regulate is best described as nascent. The technology is employed largely for back-office functions, with customer-facing technology largely in the exploration stage.”
Some of the most useful applications of artificial intelligence and machine learning may be quite humble. Take, for example, the potential to better manage the ever-expanding volumes of work emails. Over 124 billion business emails were sent worldwide each day in 2018, and that is projected to rise to over 129 billion each day by the end of this year, according to Ludré Stevens, co-founder and chief product officer of Inbotiqa, a fintech that aims to help banks substantially improve individual, team and company productivity and customer service levels, while dramatically reducing operational risk and costs – all through more intelligent use of email.
Ludré Stevens, Inbotiqa
Stevens has 20 years’ experience in banking, much of it spent running large technology and process-change projects at firms such as Morgan Stanley and Lehman Brothers. He founded Inbotiqa in 2011 after seeing a problem at another firm.
“It lost $2 million because it didn’t see a single email,” Stevens tells Euromoney. “Email is universally used by all businesses, by their customers and suppliers. It is an official communications tool with great flexibility to attach documents, for example. But it was never designed for business. It just came to be used by every business.”
He continues: “What many firms need, especially businesses using shared mailboxes receiving high volumes of incoming email, is better tools to track and manage those emails – who they have been allocated to, how quickly and completely they have been dealt with, what are the turnaround times – to prioritize them, time stamp them and to establish full audit trails.”
Financial services companies used to be good at tracking the volumes of phone calls sales teams made and their resulting success rates, but don’t have similar tools for email. This is surprising, given how frequently email is used for the essential plumbing of financial services, for example in middle-office and back-office operations functions.
Inbotiqa applies a basic workflow tool on top of email, which does not require senders and recipients to alter in any way what they put in their messages to improve the function. The firm, which came through Barclays Techstars accelerator programme, has a number of tier-one banks as clients, though for now it can only identify ING.
Email is universally used by all businesses, by their customers and suppliers. It is an official communications tool ... But it was never designed for business- Ludré Stevens, Inbotiqa
Stevens talks through a recent pilot for another bank dealing with high volumes of settlement instructions sent into shared mailboxes rather than to named individuals. “This bank had multiple shared inboxes in different regions of the world, handling settlement instructions in various underlying asset classes, including equity, fixed income, foreign exchange. In some cases, counterparties would send the same instructions to multiple inboxes.
“The bank started a pilot with our YUDOmail product in a specific asset class. In the first month, that bank was turning around only 21% of these incoming emails on the same day they arrived. It was spending a lot of time chasing unresolved settlement instructions and trade breaks. Remember: every trade break has balance sheet implications – I still have the client’s money or the client has mine – and that has costs and impacts on client satisfaction.
“By the third month of the pilot, that bank was turning around 84% of incoming emails on the same day they were received.”
He adds: “This reduces operational risk. The tool records who an email has been assigned to, whether and for how long it remains unopened, if it has been read and whether it been dealt with. It also helps identify bottlenecks. The bank might spot that a lot of trade breaks originate with the same counterparty that might be inputting trade identifiers incorrectly or missing out key settlement instructions.”
Liza Russell, chief executive of Inbotiqa, who made her career in operations in retail banking and wealth management at RBS and Coutts before a stint at PwC, explains how the firm’s next product takes this up a level. “YUDOsmart’s natural language processing can detect changes in the tone of emails – ‘we have repeatedly queried this, it’s urgent that we resolve this’ – and will help firms prioritize tasks and at the same time balance the load of work, by spotting bottlenecks and allocating to people or departments with the capacity to handle tasks,” she says.
YUDOsmart identifies what tasks emails relate to and assigns them to the correct recipient, in part by tracking the path of previous emails with similar content. It could collapse multiple different shared mailboxes for settlement instruction to just one and then route them to the teams handling settlement in the right location and asset class. It also gathers together related attachments, making them easier to find and review.
Liza Russell, Inbotiqa
Russell adds: “One of the features we have in YUDOmail, which our customers love, is a ‘get next’ button, where back-office staff that have worked through their existing tasks can get the next piece of work coming in and show their bosses they are top performers. Banks are good at working out incentives for this kind of behaviour. And we can monitor that they are completing their tasks, not just opening emails, looking at them briefly and forwarding them to someone else to resolve.”
So high are the expectations around artificial intelligence that some banks initially think Inbotiqa is offering a new tool for trade matching. It is not.
“We won’t do the work for you,” says Stevens. “We won’t match up settlement instructions. We are the routing partner, not the executing partner. But we will route the email to the right place. Doing so brings an awful lot of helpful structure to previously unstructured workflow. And that helps reduce the volume of email. The bank on the pilot I described reduced email volume by 43%. Meanwhile, the associated metadata might reveal previously unspotted patterns that can improve workflow management and efficiency.”
Inbotiqa has started in banking and hopes that many financial institutions will take its service, so that it becomes something of an industry-wide infrastructure play. But it has obvious applications for other industries that feature high-volume shared mailboxes, including phone companies, utilities and government departments. It connects to clients’ own email servers, so doesn’t impact on cyber security, rather operating behind each clients’ own defences. The revenue model is to charge a licence fee per user per month.
Headquartered in London, with some operations in India, Inbotiqa is looking to further build out its tech hub in Belfast. It is in fund-raising mode, talking to venture capitalists and banks about a pre-series A funding round.