Back to Blogs

Solving Unique Legal Problems Using Machine Learning and Expert Teams

January 16, 2020

contracts ai machine learning technology and digital contraxsuite ai and data science daniel katz libor

There is a lot of buzz regarding the change AI is bringing to the legal space–who hasn’t read an article about “robot lawyers” coming to take our jobs? On some level we know this isn’t an accurate forecast, but the media thrives on the vagueness and uncertainty surrounding AI. Meanwhile, it’s often difficult for GCs to determine if a software pitch is the right solution for their legal operation needs.

Some of this interpretative struggle is due to the seemingly endless applications for legal tech solutions. That’s why it is critical to understand that AI brings new processes to the table, but that lawyers and legal professionals will always work in tandem with AI. It’s not humans or machines; it’s “Humans + Machines.”

What is LIBOR?

In July 2017, the UK’s Financial Conduct Authority (FCA) announced that by the end of 2021 British banks would no longer be required to submit rates for the London Interbank Offered Rate (LIBOR). This means that LIBOR will play a diminished role in the global financial system going forward, and may disappear altogether.

Ever since that announcement, law firms, corporate legal departments, ALSPs, and everyone in between have been asking: How many of our clients have “LIBOR-infected” contracts? How much money is at stake? How do we prioritize and facilitate re-negotiation and re-papering?

The first hurdle is to identify the LIBOR-infected contracts. After that, teams of lawyers and legal professionals can work on remediation. That’s a two-pronged problem, requiring a two-pronged solution.

Identifying LIBOR Contracts with Machine Learning

Discussing machine learning (ML) solutions brings us back to that “robot lawyers” misunderstanding. A well-trained ML platform will find a lot of relevant data points in a large set of documents, and can be adapted and customized with additional ML techniques to meet the unique challenges posed by the LIBOR problem. Designing and building these robust techniques requires strategic planning and communication between various types of subject matter experts. A software development team can build ML algorithms in multiple different ways, but it takes experts in law and finance to fully flesh out all they need from a custom LIBOR analysis tool. Basically, “robot lawyers” don’t exist; effective ML requires a range of human experts to sit down and discuss how best to solve incredibly complex problems in a sophisticated and results-oriented way.

ML is not a conveyor belt where data goes in and perfect results come out. It takes time and iteration. This is actually what makes ML a natural companion discipline for legal: both disciplines require taking in imperfect data, then developing creative and effective solutions with that data.

For example, an ML implementation team handling “LIBOR-infected” contracts must ask whether a few natural language processing (NLP) techniques can find all the required data, or whether more complex vectorization models are needed. Data points are neither simple nor intuitive: spread percentages; governing law clauses; jurisdiction-specific legal language; synonymous or nearly-synonymous terms such as “Eurodollar”; and fallback clauses tied to other reference rates such as SOFR and SONIA.

ML can solve a lot of problems in legal, but sometimes it’s forgotten just how vital the contributions by experts are. At the end of the day, ML is just complex software. ML is only as good as the team that builds it, oversees it, and shepherds its evolution.

Solving LIBOR Remediation with Expert Services

The LIBOR problem requires teams of various specialists. Many organizations already have such teams, or at least a set of processes in place. Service teams from law companies like Elevate must be nimble enough to integrate with clients and their processes in order to augment what is already there. A client may not want to use ML, but an outside services team may recognize the potential for ML deployment or see that it is wiser to simply ramp up the human review team.

Either way, for an acute problem such as LIBOR, a services team must bring necessary resources, processes, and technology to their client’s team, and help deliver efficiency and cost savings in four major ways: quantification, action planning, remediation, and reporting.

Quantification: To properly determine the level of repapering a client needs to remain compliant and reduce risk, any outside services team should know how to quantify their client’s LIBOR exposure with a full historical assessment of the contract paper in question.

This initial quantification is effort-intensive, which is why many organizations bring in the efficient expertise found in law companies. Moving fast and iterating is just as important for AI developer teams as it is for a specialized services team.

Quantification efforts will typically include:

  • Scoping
  • Targeting data repositories for relevant contract data
  • Identifying the contracts most impacted by LIBOR (perhaps using a platform like ContraxSuite)
  • Reviewing and summarizing contract information for further analysis
  • Working with a client’s pre-existing methodologies to provide internal stakeholders with the clearest picture: which agreements are impacted, their level of exposure, and actions needed

Action Planning: After quantification comes the action plan. For a services team, this may mean:

  • Supporting Legal Project Managers as they help their teams map out necessary steps and delegate tasks
  • Coordinating parties for effort estimates and accountability
  • Identifying, assisting, and leveraging third parties (e.g., outside counsel, law companies, technology providers, and/or experts)
  • Building consensus and internal buy-in for final action plan

Remediation: After quantification and action planning, it’s time to finalize the review and tally up the contracts that require remediation. The following are five general types of remediation:

  • Level 1: Tracking all “no action” contracts
  • Level 2: Notification (outgoing) of LIBOR rate transition
  • Level 3: Notification and simple remediation of contract (no countersignature required)
  • Level 4: Notification and simple remediation of contract (countersignature required)
  • Level 5: Notification and full remediation of contract

To assist with the facilitation and resolution of these items, organizations will most likely need to hire temporary contractors or outsource the remediation process to a legal service provider.

Reporting: Throughout the process, it is important for legal teams to be aware of the progress being made, agreements pending, agreements remediated, cycle time, type of remediation, etc. A good services team knows how to support the management, tracking, and presentation of reports for internal stakeholders, ensuring accuracy on scope, quality, and budget over time. Again, the use of AI tools by a services team can take this even further, providing deeper data- based insights for future projects.


“Humans + Machines” is better than humans or machines by themselves. We have delivered contract insights on existing and remediated agreements, using the specialized skills of our services teams, and powerful software tools like contract analytics platforms that sift through thousands of LIBOR-infected contracts. The current centrality of the LIBOR problem is just one of the many examples of “Humans + Machines” completing high-quality enterprise legal work. It will only get better from here.

Machine Learning (ML) is not a conveyor belt where data goes in and perfect results come out. It takes time and iteration. This is actually what makes ML a natural companion discipline for legal: both disciplines require taking in imperfect data, then developing creative and effective solutions with that data.

Back to Blogs