The date looms large: 12-31-2021. Barring a delay, New Year’s Eve 2021 marks the last day that the Libor benchmark is to be published. While that’s the date pegged by the Financial Conduct Authority (FCA), the ostensible death knell for the discredited rate could actually come sooner than that.

There are currently more questions than answers for exactly how this transition to an alternative benchmark—such as the Secured Overnight Financing Rate (Sofr) for US dollar contracts or the Sterling Overnight Index Average (Sonia) for the British pound—will unfold. Numerix isn’t sitting back and waiting for these questions to unfold at what feels like a glacial pace.

The risk technology provider has teamed up with NextGen Strategic Advisors to introduce a new module, dubbed Oneview for the Libor Transition. The module aims to help firms (mainly those outside of the top 10 largest banks) overcome the legal, operational, technological, and risk challenges associated with Libor’s discontinuation.

While this may seem like a legal challenge and a “paper” challenge, as old contracts need to be renegotiated—and, to be sure, it is—this is also a massive technological haul, says Steven O’Hanlon, Numerix’s CEO. First, a bank will have to locate these contracts, as many of them are not digitized and are off-premises, he says. After that, the lawyers come in to review all those legal documents, find all the language pertaining to Libor, and then set forth addendums for replacement language around the alternative reference rates.

“That’s no easy task,” O’Hanlon says. “Many clients who we’ve been talking to say it could cost as much $800 per contract to get junior lawyers to attack this problem.”

Last year, Numerix added Sofr and Sonia curve-building support through its CrossAsset Analytics platform. To price and hedge interest rate derivatives, which are often very long-dated, users need an accurate interest rate curve. Numerix offers curve instruments and a curve framework for Sofr and Sonia.

That project served as the basis for the Libor Transition module. For this latest leg, Numerix brought on NextGen Strategic Advisors and its managing partner Gary Mandelblatt, who worked alongside Ron Coleman, who was previously at IBM in the Supercomputing Systems Lab, where he was part of a team that built a version of what would become Deep Blue, the chess-playing supercomputer. Coleman provided his expertise in machine learning and natural language processing (NLP), and that expertise was combined with Mandelblatt’s knowledge of financial contracts, Libor, and the various alternative rates.

The Libor Transition module, which was written in Python, uses an open-sourced NLP tool to read documents and pluck out the necessary legal terms, and Google’s open-sourced TensorFlow for the machine-learning component to provide a modeled curve structure and volatility surface—volatilities used to price trade instruments—as well as to flow documents either into a bank’s internal systems or into Numerix’s Oneview platform. The AI can also digitize a contract where needed to make it optical character recognition (OCR) compliant.

“Once we’ve done the AI work to expose what the new rates would be, then the managing directors of the bank go out and talk to their clients, negotiate it, come back with the rate, and we can take that file and push it into the bank’s own system. Or, they can push it into our Oneview risk system, which has a Libor module now that gives the ability to containerize all of that and set forth a way to look at the rate into the future and actually benchmark back and forth,” O’Hanlon says.

Additionally, as more contracts are fed into the system, the AI produces a “percentage of accuracy” based on how it’s taught.

“Once you reach an accuracy [rate] of 98%, it’s pretty safe to assume now that it’s learned enough that a lawyer can do a quick scan of 10 or 15 of these and recognize that it’s spot on and there’s no further training [necessary],” he says.

A Tech Issue

NextGen’s Mandelblatt says that this transition process starts off “in paper” and quickly becomes a legal issue, which then ties into operations and technology. He notes that this is also a risk management event.

“The technology component of this is huge,” he says. “Once we get the legal documents updated and we know what the new rules are, we have to get that into a system and we also have to perform risk management sensitivity analysis on these new curves because they are all new—we don’t know what the forward curve structure looks like from experience; we can only model it out.”

That’s just the beginning—what happens to a volatility surface, because there isn’t one for, say, Sofr? The legal language around these contracts transforms legal risk into operational risk, he says, so that information needs to move seamlessly through the system after it’s been hammered out on a legal level.

So the tech has to be able to identify what kind of product is being addressed in the document, how it’s being impacted, what the new rate is, and what the volatility is during different points of the volatility surface. That only becomes more challenging for complex instruments, such as swaptions.

“This is a story that goes rapidly from legal, into operations, into risk management, which takes you to valuation trading, and technology is the core of all of that in being able to measure it,” Mandelblatt says. “A financial contract is just a piece of paper until there’s a model, and getting the models to work right is what’s so hard in this effort. It’s more than just going from one curve to another—it’s that these curves aren’t understood. So you can turn it on, but you have to get it right.”

The challenge only gets more convoluted when you factor in contracts written pre-October 2017 when the FCA announced the discontinuation of Libor. Banks need to figure out which contracts have pre-cessation fallback triggers for derivatives transactions that use Libor as the reference rate; while those contracts might’ve had language around Libor not being available, they did not address Libor going away altogether.

“A lot of this information is not built into any system because we never anticipated Libor going away,” he says.