The Jetsons, the 1960’s animated sitcom, presents a particularly optimistic view of automation. In the year 2062, salaryman George Jetson holds down a full-time job of two hours a week. Everything else having been automated, George focuses solely on doing what he is uniquely capable of doing. (Spoiler: he pushes a button up to five times a week)
There’s a lesson for us here in transitioning contracts from IBORs to Risk-Free Rates. The sheer scale of the upfront contract review and assessment effort alone begs a technology solution, so we can focus human judgment where it is most needed. While we shouldn’t expect to meet George Jetson levels of automation, technology adoption can make LIBOR projects far more efficient and save countless hours of manual drudgery.
The contract review problem for Libor transition is unprecedented: our research estimates that up to 100 million contracts need to be assessed. Technology should play a vital role in reducing the overhead here.
At this stage, you are probably thinking about using AI review tools, such as Kira, Seal, or Eigen. Last year, we reassessed the market for these tools, tested the market leaders, and redesigned our review processes. This was part of a broader, year-long project to prepare for Libor transition. This preparation and, before that, a decade of delivering large-scale, complex, technology-enabled contract review and remediation projects, have informed our views.
Here’s what we have learned: summarized in three fundamental principles:
Let’s look at these three principles and how they apply specifically to contract review and assessment for Libor transition.
All technology is a means to an end. Rather than starting with specific software and vendors, begin by thinking about where you want to be at the end of your Libor program. From here, systematically work back to understand what you will need at each stage to achieve those objectives.
Performing the right amount of review is crucial. Unnecessary review, even using technology, is costly. Too little review can lead to rework, when you realize there is something you failed to capture the first time.
By starting with outcomes, you realize that merely finding if contracts contain relevant IBORs will be inadequate. You need to consider the information to construct amendments and to drive remediation prioritization matrices.
You may also think about what else you can extract here, while contracts are open, to drive long term benefits. For example, if you want to use a new contract lifecycle tool, what would you want from these contracts?
Bringing new technology into any existing operation evokes Bill Gates’s rules of automation:
The first rule of any technology used in a business is that automation applied to an efficient operation will magnify the efficiency. The second is that automation applied to an inefficient operation will magnify the inefficiency.
Define your process. To understand how contract review technology will fit into your project, you will need to look at your existing process. For example, where are your documents currently, who would review them, where would they need to go once the review is complete?
Measure your baseline. This can be as simple as timing how long it currently takes you to do a contract review. Early on in our Libor transition preparations, we baselined by testing how long manual review took and then compared how long it took reviewers to produce the same output with trained AI. These measurements not only helped us understand uplift but allowed us to identify unexpected bottlenecks. For example, team members unfamiliar with AI platforms sometimes produced slower results with AI than they did manually. This insight revealed that they didn’t know how to spot signs that the AI was performing well versus underperforming, and we were able to provide training around this.
Be holistic in measuring this baseline. Whether you’re using AI or performing a manual review, a lot of the upfront work will be the same. You will have to find contract files and give them some essential data attributes to be able to track them.
Remember to also baseline the AI’s performance. Some vendors have done a great job of building out of the box algorithms for Libor transition but remember that these were created on publicly available data sets, across just a few (of the many) product areas impacted by IBOR transition. For other asset classes, where your contracts are sufficiently different to the training set, or where you want additional data points, you will need to do some training.
Baseline the accuracy of an existing trained algorithm against a sample of your documents. Where the performance of your algorithm isn’t satisfactory, or you need a wholly new algorithm, you will need to perform training on an algorithm. Consider, given your manual baseline and the volume of documents, if this will be efficient for your entire population, a particular contract type, or a specific contract term.
Design for your standard. Although a wide variety of contracts are impacted by Libor transition, there are common elements across these that can be defined. You will have to carry out the same upfront work to gather these contracts and prepare them for review. Review itself can be standardized based on the routes to remediation you expect the different contract types to take.
Particularly where you are utilizing a large team, designing for the interactions they have with the technology and between team members is essential. Consider how your team will review the output of the AI tools and how senior team members will support the work of junior team members.
The delta between the data you have and the information you need is where you should focus efforts. This includes thinking about the reports you will need to drive decisions and to provide visibility to the project, across jurisdictions, business lines, and stakeholders. For example, you will want to track the progress of linked products (e.g. loans and hedging instruments) together rather than in isolation.
Begin by creating a data model of all the information that you will need about your contracts for Libor remediation, including the downstream data that you will need for amendment and prioritization matrices, considering the exact format in which you need that data. Now map the information you have in your existing systems, taking care to consider if this data is of a quality on which you can rely. The difference between these two states will be where you need to focus your Libor contract review.
Also, consider how you want to be able to report this information from contract review. As we progress towards the end of 2021, the data that you find in review will be vital for your organization’s decision making and for communicating your exposure and progress to regulators and other stakeholders.
The end of 2021 is fast approaching, and while we’re still waiting for flying cars, as far as Libor is concerned, we can always take a lesson from the Jetsons and let the machine do (some of) the work.
