How the use of technology is changing due diligence processes in portfolio sales

In brief...

Technology is evolving at a rapid rate within the legal sector, and is increasingly affecting the way in which deals are conducted. Law firms have found particular applications for augmented intelligence technologies in portfolio transactions, where machine learning software can assist with large-scale document review and similar due-diligence processes. This article considers how machine learning software is changing these processes, and discusses the potential value that it can add.

Artificial intelligence (AI) and the ground-breaking Internet of Things (IoT) continue to hit headlines. Back in 2014, Stephen Hawking stated that success in the creation of `true' AI could spell the end of the human race. However, while AI is popularly portrayed as the dawn of human-like robots, those in the industry often prefer the term "augmented intelligence" for machinelearning technologies which are not meant to replicate full humandecision making, and probably will not for some time.

Law firms are increasingly using such augmented intelligence technologies to improve and change the way in which they conduct deals, such as portfolio transactions. The licence agreement between DLA Piper and Canadian technology firm Kira Systems (Kira) is one such example of our ventures into the world of augmented intelligence. Kira is a contract review and data extraction tool. It `learns' to accurately identify provisions in large numbers of documents and rapidly presents them to the reviewer on a user-friendly platform - allowing faster document analysis by lawyers.

The transfer of large volumes of loans or other financial assets will often require extensive document review by law firms acting for both vendors and purchasers. This article considers how machine learning could change this process.

How does augmented intelligence add value?

Augmented intelligence software allows the rapid review of large volumes of data to reveal patterns, trends and interrelationships. The performance of this software improves with exposure to data; the more transaction data it processes, the more accurate the software becomes, and the better risks can be mitigated. Kira, for example, can accurately identify information by learning from examples; this is in stark comparison to its predecessors, which could only identify pre-programmed clauses. Indeed, technology is now sufficiently advanced that it is possible for people with limited software-specific knowledge to train Kira and other such software to meet particular project needs.

To fully determine the benefits that machine learning can bring, it is necessary to consider the current method of reviewing large volumes of documents. Traditionally, transactions which involved the sale of large portfolios of loans (as part of a securitisation or a more conventional sale) have, to varying extents, required a review of the underlying contracts which documented those assets. Review of these documents is inevitably a time-consuming, and accordingly, costly process (even with the efficiencies achieved by employing junior lawyers).

There is a clear advantage in deploying software to review, analyse and report on the contents and key terms of thousands of documents within minutes or hours (rather than a manual review taking days or even weeks).

As well as increased efficiencies, machine-learning software may provide law firms with an opportunity to conduct document review on a far larger scale than has previously been practicable. Constraints of time and cost have meant that traditionally, some reviews have adopted a sampling approach, assuming a degree of uniformity among documents and assessing risk accordingly. In theory at least, document review software now makes more comprehensive large-scale review achievable. Moreover, developers and users of this software claim it limits the margin for error that any human review may present.

Are there any limitations?

While further technological advances may present opportunities for efficiencies and the scope of reviews to be expanded, it seems that there are still limitations. Crucially, machine learning in this context needs lawyers to teach the relevant technology to identify the relevant provisions. This takes time and resources. Moreover, a machine can only highlight provisions as being problematic or concerning to the extent that it has been taught to recognise them as such. Accordingly, without the fall back of a human double-checking (which may reduce efficiencies), document review software could overlook problematic or onerous document terms. It is here that flexibility in software development combined with a focus on legal input and management are vital to ensure best practice and accuracy.

Conclusion

Managed well, machine learning will make document review faster, more efficient and more economic - all of which is of benefit to clients. However, law firms still need to support software setup and provide significant and ongoing legal input to ensure that processes account for the evolving issues and risks that businesses face. So far, augmented intelligence remains a tool rather than a replacement for lawyers.