AI and LIBOR
How Practical AI Can Accelerate Your Transition Away From LIBOR
by Peter Wallqvist, VP of Strategy at iManage RAVN
For years the London Interbank Offered Rate (LIBOR) has been the world’s most widely adopted interest rate benchmark, leading to its use as a base rate for financial contracts worth more than $300 trillion worldwide. This includes trillions of dollars’ worth of U.S. financial contracts – including, according to the U.S. Alternative Reference Rates Committee, $1.2 trillion of U.S. retail mortgages alone.
Alas, support for the LIBOR benchmark will be phased out by 2021. As a result, any organization with financial contracts using LIBOR needs to transition these contracts to alternative rates, such as the Sterling Over Night Index Average (SONIA) or the Secured Overnight Financing Rate (SOFR), or risk finding interest rates in these documents linked to a benchmark that no longer exists.
How can organizations identify which of their thousands of contracts need to be transitioned away from LIBOR, as well as where and in what ways these contracts need to be updated?
Organizations could try to manually tackle this task with a team of lawyers, paralegals, or other professionals, but this work would likely require years of man-hours and cost millions of dollars, with significant potential for human error – an unpleasant proposition any way you slice it.
The daunting nature of the task at hand is enough to make any financial institution throw up their hands in despair. Given the rapidly approaching 2021 deadline, however, inaction is not a viable strategy – the clock is ticking.
A Practical Approach to an Enormous Undertaking
Fortunately, new practical AI technologies have emerged that provide organizations with an alternative approach.
When we speak about “practical AI,” we’re referring to products and solutions that use artificial intelligence to automate routine work and tasks, address common pain points, and solve real-world problems. With practical AI, organizations can automate the identification of contracts that need to be repapered for the transition away from LIBOR.
The first step in the process is digitization, which ensures that the data in the financial contracts is actually “machine readable.” After all, AI can’t be expected to do much with a box of paper contracts.
Once digitized, it becomes possible for AI products to analyze the contracts, understand the presence or absence of key clauses or other pieces of data, and then use “decision tree” logic to determine next steps.
For example, let’s say you start with a pile of 40,000 contracts. AI will analyze those contracts and extract the “Termination Date” for each one to identify which ones are expiring before 2021. If it turns out that 60% of those contracts expire before 2021, congratulations: those items can be disregarded because they won’t be impacted by the LIBOR transition, and you now have a much smaller pile.
Next, AI will analyze the contracts that continue through 2021 and beyond to determine if the interest rate in the contract is based on LIBOR. If not, congratulations again: that’s another pile of documents you can disregard.
If the interest rate in the contract is based on LIBOR, AI will look to another data point to determine next steps: is there a fall-back provision in the contract for situations when LIBOR is unavailable? If not, the contract certainly needs to be amended.
If there isa fall-back provision, AI will analyze the provision to determine whether it’s broad enough to deal with the permanent discontinuation of LIBOR as an industry standard as opposed to a temporary unavailability. If the provision is sufficiently broad or inclusive, then congratulations once more: the contract will not require amendments. If it isn’t, AI will flag the contract for repapering.
In this way – one step of the decision tree at a time – AI can whittle down an immense mountain of contracts to a much more manageable number, helping firms identify which contracts need to be amended in preparation for the transition away from LIBOR.
The Human-Machine Partnership
It’s important to note here that humans remain very much involved throughout this entire process, operating in a training and oversight capacity to ensure AI is delivering optimum results and the highest levels of accuracy.This makes practical AI the epitome of a human-machine partnership, where AI handles the tedious “grunt work,” while humans provide ongoing training and guidance.
Better yet, once organizations have trained their machine learning models for this specific LIBOR-related task, they can reuse them in the future for other tasks – helping to maximize return on their investments in these practical AI solutions.
In an ever shifting legal, regulatory, and economic landscape – Brexit, anyone? – the need to amend large numbers of contracts will be the rule rather than the exception. With its ability to quickly analyze and interpret scores of documents, practical AI can help organizations meet the challenge head on.
Technology has enhanced capabilities of surveilling larger and more disparate data sets.
With Eugene Kanevsky, James Redbourn, and Joanna Wong, CLSA
AI/ML on the buy-side trading desk is a long-term program with short- and mid-term deliverables.
Aiden uses the computational power of deep reinforcement learning to improve trading results and insights.
BondDroid’s AI-generated prices are integrated directly into LTX’s pre-trade analytical tools.