30 June 2023
When the big world-changing events of 2023 are written down (inevitably by a computer), we can be extremely confident that the acceleration of Artificial Intelligence (AI) innovation will be one of the year’s dominant headlines.
At this stage it’s hard to know how transformative the technology will eventually be – nonetheless use cases are emerging in almost every sector of the economy. In the financial services sector AI is already proving useful in some areas, such as fraud prevention and risk management, while being challenged in others, particularly where data sets are limited.
But does it have what it takes to power the vital lifecycle of loans documentation?
Answering this question is anything but straightforward. Yes, in theory AI-powered platforms should be able to assist with contract analysis, quickly and accurately reviewing legal documents to identify key terms, provisions and potential risks.
Additionally, there should be lots of potential for AI to manage contracts throughout their lifecycle by automating tasks such as contract creation, renewal and termination, while drawing attention to any breaches.
Envisaging AI’s potential capabilities is much easier than applying the technology in the real world, especially when it comes down to loan documentation:
1_ Data quality
AI models require high-quality and high-volume data to produce accurate results. Loans documentation has a huge degree of legal language variation and the documentation can be complex and contain inconsistencies, which will almost certainly affect the accuracy of AI models.
2_ Human intervention
Due to this variation and complexity in loans, AI will embed more (or unlikely remove) manual processes into firms’ operating models as a human will always need to validate the answer the machine gives you for ‘false positives’.
3_ Bias and fairness
AI models can inadvertently perpetuate bias and discrimination if they are trained on biassed data or use biassed algorithms. This could result in unfair lending practices potentially breaching regulations relating to customer service.
4_ Transparency and explainability
AI models can be difficult to interpret and understand, which can make it challenging to explain the reasoning behind lending decisions to customers or regulators – the dreaded ‘black box’.
5_ Security and privacy
Loans documentation contains sensitive personal and financial information, so it is essential to ensure that AI systems are secure and protect customer privacy.
There are also a range of specific and difficult functional challenges found in loan agreements that AI technologies in this space have yet to master, such as:
Another thing to bear in mind with AI is regulation. Like everyone else, governments around the world are racing to understand and manage the impacts of AI. From the European Union to China, new regulations are being drawn up and debated by experts and policymakers.
But like the technology itself, the implementation of workable rules is sure to be a complex and painful process, particularly in the financial services arena, which is already heavily regulated.
We’ve seen how AI can be beneficial in some areas and struggle in others. AI requires a much greater degree of human intervention and we have seen this lead to diminished ROI, and problems with delivering against the business case.
Likezero’s data-mining and matching platform is able to automate the analysis of a wide range of data in legal documentation without manual inputs. AI cannot compete here, because it requires a much higher volume of data to train before it can deliver a similar performance. If your portfolio has a limited data set, AI simply won’t work.
Likezero’ solution uses unique data mining and language matching to extract legal terms and data. The result is precision and ends the need for human review for any data the software has collected.
See what our clients have to say below:
Head of Portfolio Management, Global Investment Bank
Our approach also ensures that data models can be extended seamlessly across all contracts and that there is no need to ask developers to code or maintain rules.
The team at Likezero has deep experience in previously leveraging AI and machine learning for legal documentation, having worked in this market for more than a decade.
We know that AI and machine learning are already adding value in fraud detection and basic data capture from vendor agreements. However, in areas such as standardisation and driving efficiency in origination to achieve legal optimisation, portfolio analysis and comprehensive data capture to feed downstream systems, we know that our technology is – and will always be – much more capable.
In our opinion, while we need to embrace AI it’s vital to recognise its limitations. Otherwise we might find that the year of AI acceleration will turn out to be something of a false start.
