With references to the soon-to-be-phased-out benchmark across thousands of contracts, banks have enlisted artificial intelligence, machine learning and natural language processing to support the transition effort.
“I think we have something in the neighbourhood of a couple of hundred different document types that could have the word Libor [London Interbank Offered Rate] in them,” observes Jason Granet, head of Goldman Sachs’s Libor transition team.
The “single biggest aid” of being able to use artificial intelligence (AI) as part of the Libor transition process, Mr Granet says, has been “to get through documents in a more efficient fashion”.
And given the scale of work that needs to be done, such efficiency gains can be substantial.
Libor is the world’s most-referenced financial benchmark. It has been used in countless financial contracts – derivatives, bonds and loans – for decades. A recent Financial Conduct Authority estimate put the value of outstanding Libor-referencing contracts at $260tn globally. But its days are numbered.
If, as expected, Libor’s administrator confirms widely-consulted plans, it will cease to be published for four out of the five currencies for which it is currently calculated at the end of 2021, and for US dollars in mid-2023.
Global regulators have been pushing market participants to transition away from the benchmark for several years, not only moving to alternative benchmarks for new issuance but ensuring vast swathes of legacy contracts no longer refer to Libor either.
Administrative burden
Given the widespread use of Libor, identifying which contracts need to be remediated has the potential to be an administrative nightmare.
Lewis Liu, CEO of Eigen Technologies, a document AI platform for financial services, says: “The sheer scale of the issue means that banks need to review pretty much every contract to check for Libor references. Some banks are choosing to do it manually, but the cost in human resources to do it that way is substantial.”
At a basic level, platforms like Eigen work by programming the technology to use a series of specific queries, which have been defined by the client, to interrogate documentation. In the case of Libor transition, such queries could be identifying which reference rate is used in a contract, what is the maturity date of the contract or even whether so-called fallback language (language that defines what happens in the event of a reference rate ceasing to be available) is present in the contract and, if so, does it adequately address the issue.
“It is possible to ask very specific questions via the platform,” Mr Liu says. He claims the Eigen platform – which has been used by Goldman Sachs and ING – is among the more sophisticated platforms on the market because it does not require large volumes of examples to teach its algorithms what to look for.
The sheer scale of the issue means that banks need to review pretty much every contract to check for Libor references.
The platform will then flag the contracts which need addressing and what likely remediation is needed. Eigen estimates that when ING used the technology in its Ibor transition programme, it cut the cost of the programme by 60% compared with manual processing and was about 75% quicker.
Deepak Sitlani, a derivatives partner at Linklaters, says: “On the data extraction side, we certainly are seeing a combination of AI to deal with easy wins, combined with good old-fashioned human resource as well.” However, he also adds that there is some variation in what different platforms can offer, saying: “The problem is that it is sometimes easy to overplay what AI can actually do.”
Mr Sitlani adds further developments on the repapering side are likely where AI could increasingly be used in the actual process of repapering contracts, once they have been identified, too. “On the repapering side, there’s going to be lots of paperwork flying around this year, so there’s been a flurry of activity on contract automation and online negotiation tools,” he says.
The right effort upfront
But AI is certainly not a silver bullet for tackling the reams of Libor legacy contracts in existence. Both Mr Liu and Mr Granet emphasise the importance of thorough upfront planning, asking the right questions and putting in place the right structures to enable the technology to be effective.
Jason Pugh, managing director of D2 Legal Technology, a legal data consulting firm, which has worked with several global systemically important banks on Libor transition, echoes this view. “It very much needs to be a combined human and machine effort,” he says.
“In many cases the technology is only as good as the initial human inputs. For it to be effective, an organisation needs to be willing to spend the time carefully defining specific questions for each different asset class and instrument. It needs structure and discipline in order to be useful.”
Mr Pugh highlights how it is more complex than simply asking the technology to go and find all references to Libor. In relation to fallback language for example, he points out the word ‘fallback’ is not often used in contracts, and other related terminology needs to be considered. For the technology to work, firms need to spend time understanding within their organisation the type of language and structures that have commonly been used in contracts for each of their product areas.
Long-term benefits
Although the technology can have a powerful impact in relation to Libor, given that the transition requires banks to process a substantial proportion of all their contracts, there is also the potential for a much bigger prize. Some organisations are opting to use this as an opportunity to create well-structured digital documentation and associated workflows, which will enable documents to be managed far more efficiently across the board in the future, rather than viewing this as a one-off exercise.
Mr Granet says: “One of the main reasons why the technology has been a value-add is because it can help you digitise your documents for the long term. A lot of engineering time is required on the front and back end to teach the system, but you also get a lasting digital footprint, which can make a difference for the future.”
Eigen’s Mr Liu also points out that once a client has done the work of scrutinising contracts for Libor-related issues, it is then possible to introduce “many more questions” that can be used interrogate documents in relation to a whole host of other areas such as “risk-weighted assets optimisation or in relation to ESG [environment, social and governance] criteria”.
This is a view shared by Mr Pugh, who argues that Libor transition will be a significant catalyst for the growth and development of so-called ‘legaltech’, with banks putting in place systems that will allow them to respond to future regulatory changes far more efficiently and painlessly.
“Since the financial crisis there have been a number of different regulatory reforms such as QFC [Qualified Financial Contract] rules in the US or the recovery and resolution regime in Europe,” Mr Pugh says. “What organisations have typically done is review the relevant documentation, identified the relevant clause and then amended it. But then, guess what? Another regulatory development happens.”
He says companies which take a strategic, rather than tactical, approach now and use the Libor phase-out as a wider opportunity to invest in digitalisation will have an operational advantage in the future. Not only will they be able to respond to regulatory changes more quickly, but with “cheaper front to back processes, better client service and less risk”.
The end of 2021 – when the transition needs to be completed in most jurisdictions – is not far into the future, so the opportunity to be strategic rather than pragmatic may be limited for those who have not already begun the process. However, for US-based banks which are only dealing with US dollar Libor, the window of opportunity may have been extended.
