Contractual Remediation : An use case for Artificial Intelligence

Execution of IBOR transition is a humongous task for any bank as interest rate is the core of any banks's business model and profitability. Those who are involved in IBOR transition across industry must be aware that one of the biggest challenges in the transition at the current stage is to go through the heap of contracts clause by clause and identify the IBOR references. Even before that locating, collating and digitization of the contracts are another troublesome areas. Renegotiation and repapering are the subsequent exercises which will bring their own set of challenges. 

Below is an effort to summaries the list of challenges associated with contract discovery and fallback identification.

Scattered and unstructured data

• Locating : Millions of documents, recklessly scanned files, handwritten memos are scattered across a complex matrix of file systems, offline storage devices and megabyte desktops.
• Variability : The divergence in the quantity and quality of the documents is immense. These contracts could be standardized master agreements, bespoke contracts of structured transactions, user agreements, consent papers, issuers circulars etc. Each line of business could have their own internal documents driving the legal clauses or underlying trades behavior.
• Volume: On one hand there could be only one issuer circular document driving the bond issuance of hundreds of dollars. On the other hand there are thousands of credit card user agreements which need to be scanned to find references to IBOR driven penalty charges. 
• Complexity : The variability of legal language, jurisdiction , currency and product specific nuances, cross references to multiple documents makes it harder to rely on a definitive set of keywords for applying machine learning solutions.

Manual Effort

Scrutinizing each document clause by clause with a legal pair of eyes is a mammoth effort which requires an army of lawyers, professional, paralegal staff the years of continuous effort. But the clock is ticking and the LIBOR demise is only 13 months away. Industry can not afford to lag behind on the documentation part at least. Furthermore the manual scrutinization has a high possibility of human error which can be quite costly for the organization. 

USE CASE for Artificial Intelligence

Key fallback Provisions in contracts were potentially never extracted, captured and stored at accessible centralized repositories. Thus now for their identification the contracts will have to be scrutinized line by line. Not only the identification but the completeness and robustness of the fallback language shall also be assessed automatically. The challenges are the multitude of risk free rates across various currencies and jurisdictions.

The blend of technologies driven by machine learning and big data concepts and executed by Artificial Intelligence techniques seems promising to deal with the problem in most efficient way. 

Technologies like Robotic Text Automation (RTA) and Natural Language Processing (NLP) can be applied to deal with the enormous load of contractual complexity. RTA and NPL enabled computational linguistics and artificial intelligence (AI) can accelerate the unstructured data processing, categorizing and translating the results in meaningful conclusions.

Pre-requisite is the digitization of the document records to make the data machine readable and searchable so that machine learning techniques can be applied.

However Applied AI alone is not enough, identification of ‘key data’ and incorporation of the key ‘words’, ‘phrases’ and ‘statements’ into the algorithms and designing the ‘decision trees’ requires crucial human involvement in the process.

Example, ‘termination date’ can be identified to classify the contract as eligible if the date is post 2021. But at the same time contracts may contain variant of other words indicating for termination date. AI can be helpful here for identification of the relevant date. If contract confirmed maturing post 2021 the next step is to analyze to identify if the contract is IBOR based or alluding to IBOR rate. Subsequently the fallback language identification and interpretation can be formalized in decision tree. Any error, misrepresentation and ambiguity in this step can incur huge economical and legal consequences for the organization.

More advanced AI applications can be trusted to modify the contracts with phrases or paragraphs of robust and acceptable fallbacks. Organizations can also try to deploy their existing, tested and functioning machine learning models to this problem. However there are many external vendors who have sophisticated and IBOR-tailored solutions which they claim to be extensively scalable and can be plugged and played with existing Contract Lifecycle Management (Contract Lifecycle Management) systems.

In the era of technological transformation it is prudent to deploy such solutions as a strategic functionality so that these suite of technologies can be leveraged for the fast changing legal and regulatory requirements and also for the contingent demands such as Brexit.

Below is the list of some of the prominent players in this space offering AI driven solutions to the IBOR specific documentation use case.

Views expressed are personal.

~ Satya