The opinions expressed here are those of the authors. They do not necessarily reflect the views or positions of UK Finance or its members.

The impact of the recent wave of technological solutions flooding into the legal sector has been far-reaching, software covering almost every aspect of the finance lawyers daily ‘to do list’, from basic proofreading through to complex drafting is now available on the market. Whilst lawyers are keen to translate these innovations into time and cost savings for clients, when it comes to legal due diligence on a secured loan portfolio sale, it is important not to overpromise - the real benefits will come from ‘humans plus AI’ rather than AI alone.

Due diligence

Legal due diligence, whether undertaken by the buyer, the seller or both, is essential when it comes to purchasing a secured portfolio.  When done, its benefits go well beyond striking the right price – a detailed knowledge of the issues within a portfolio can reduce (or even eliminate) the need for a purchase price retention mechanism, smooth the migration process between buyer and seller systems and reduce the chance that a warranty will have to be called on. It is also being inherently labour-intensive, the most promising part of a transaction when it comes to realising the benefits of AI. Loans papered on a limited number of templates, however numerous, can be quickly read and compared by the right software - and lawyers are keen to point out in any request for proposal the savings that can be made here.  They are right to do so, but it is important in promoting the benefits of AI we lawyers do not inadvertently present a picture whereby the due diligence process is reduced to an entirely automated process costing no more than the yearly licence fee on a software programme. As below, there are some things that the machine, at least for now, just cannot do.

To Err is human…and AI

First, let’s start with the obvious, AI is not yet perfect – even where all the information is available and of good quality, the large language models (LLMs) that sit behind many software options currently on the market can still make mistakes. Some of these mistakes are clear cut – for example hallucinating case law or legislation that do not exist, but some are less so, and perhaps not even fairly labelled ‘mistakes’ but more accurately, ‘product that the user does not consider right for their purposes’. AI does best when asked clearly defined questions that lead to clear cut fact based answers, unfortunately lawyers are often after the answers to complex contingent questions that tend to produce answers laden with subtlety and nuance.

Ghost inside the machine – the software won’t get it all

Many of us, particularly post the 2020 Covid pandemic, have become used to signing documents electronically. And, it is not an exaggeration to say many loan terms and conditions will never be printed.  But that is now - many loan portfolios currently in the market or yet to come to market have been papered on, well, paper. This paper, signed in wet ink (or perhaps not signed at all?), along with any manuscript amendments, was then most likely scanned and saved down electronically, with no guarantees as to the quality or readability of the scan.  Current software is good, but it is understandably limited by the quality of the input - misaligned headers, pencilled manuscript amendments or just plain illegible pages in these scans can all result in confused output.  This confusion, often in the very areas which count most when it comes to due diligence – deviation - will invariably have to be addressed by a person.

Ghost outside the machine – the software can’t process what’s not there

It seems like an obvious statement but is often overlooked – if you don’t have the underlying document, the machine can’t help you. Many older portfolios will contain loans that have been subject to amendments, rectifications or other investigations often requiring a physical copy to be brought up from archives and dispatched to internal teams or external advisors and may not be readily available. Here of course, a person cannot read it either, but that person will still be needed to formulate other mechanisms in the sale and purchase agreement to address the risk these missing documents pose, whether through representations and warranties or otherwise.

Looking to the future

This is of course not meant to pour cold water on what is undoubtedly a huge opportunity for the legal sector to reduce client costs when it comes to due diligence, but it is meant to highlight that Humans + AI invariably work better together than either on their own. AI can certainly speed up the process for lawyers on routine work, making them more efficient and lowering cost, it can also act as a second pair of eyes, reducing risk - but it is not AI that will replace lawyers - it is lawyers that use AI that will replace lawyers that don’t.