AI and M&A: a breakdown of the challenges and benefits

When it comes to using AI during the mergers and acquisitions (M&A) process there are many benefits, yet a few challenges that come to mind. In this article we investigate how these points can affect the due diligence (DD) stage for M&A transactions.

The benefits

Share capital

An M&A process, such as the sale of the entire issued share capital of a company, involves the collection and analysis of data in documents across all areas of the target company. This includes corporate structure, real estate, employment, IT and IP, commercial and litigation. It provides the ‘big picture’ of a target company and with that comes a high degree of folder structuring and data organisation. AI tools can, instead of dragging and dropping files into different locations within a virtual data room (VDR), organise information in bulk as it is uploaded and structure the VDR to make it easier and quicker to navigate. This is something that can take individuals (even with existing sorting tools) several hours, even several days, to arrange – depending on the nature and size of the transaction. 

If the share sale involves an auction process, after the data in the VDR has been organised there are various additional stages to consider. This includes the bidders marking up a share purchase agreement or being granted access to certain areas of the VDR. This depends on whether the transaction is at the initial stages or bidders have been shortlisted to the final potential buyers. 

AI tools can be used to assist sellers in providing insights into bidder motives and giving insight into which bidders are the most engaged, as it could align their behaviour to recognised patterns of successful bidders using several attributes. It’s easy to imagine professional VDR providers having this kind of data at their disposal and seeking to utilise it with AI. Although we can review market data and use our knowledge of potential buyers and their working patterns – this isn’t something that lawyers alone could provide the same degree of advice on to seller clients, without spending significantly more time. It could pave the way for greater understanding at the initial deal stages, in analysing which bidders to progress with to the next stage of due diligence (DD) and improving service delivery by saving time and costs.

Locate, create and transfer

The target company in a M&A transaction may also be part of a wider group structure, with subsidiaries that fall under different jurisdictions. We could use AI to:

• locate the various governing law clauses of documents that have been placed in a VDR
• create a list of legal advisors in overseas jurisdictions that we may need to liaise with for additional advice
• proceed with getting information to such overseas advisors at a much earlier stage of the transaction

This would all enable the parties to complete the DD exercise in better time, to focus on finalising negotiations of the main transaction documents. 

During a merger

Another example of AI assisting in a M&A transaction is during a merger - when a merger notification is drafted and sent to the Competition & Markets Authority, using information provided at DD stage. The notification includes a number of supporting documents and details of how the merger may be caught by the UK rules. An AI large language model (LLM) tool can assist this heavily-administrative task. It can help by providing answers to natural language questions, having searched across a repository of documents in the VDR, to find all the necessary references to management accounts and recent business plans. We could also ask the AI tool to “show me all references to management accounts in the finance documents”. As well as providing us with answers at great speed, the AI system is also learning from the interaction it has scanning the necessary documents, learning more quickly than humans can. 

The challenges

Quality searches

One of the primary concerns with current AI products relates to the AI’s ability to contextualise its advice or findings and offer truly commercial advice. At the moment, the output of AI tools is dependent on the quality of the input – think of ChatGPT and how its ability to provide helpful responses is subject to the amount of detail you put into your request. This isn’t too much of an issue for informal use, but when applied to complex and technical fields this can lead to risk for users. 

A large part of the benefit of using professional advisors stems from their experience of the market of similar transactions, of commonly negotiated points, and on their ability to contextualise all of this based on the needs and emotions of the parties. Linked to this, it’s often the case that clients, even those with previous transactional experience and a good grasp of the process, don’t quite know which questions they should be asking, or don’t know how they should phrase a particular question. 

AI is still a fair distance away from being able to pre-empt and form the questions users may want to ask, and also to add wider context to those questions in order to arrive at fully formed and commercially useful advice. AI output will therefore still require significant amounts of human moderation. However, it remains easy to see how AI could be utilised as a very helpful and powerful research and support tool for professional advisors. 

Accuracy

Although AI aims to be incredibly accurate and time-efficient, at the moment we can’t fully rely on its degree of accuracy. Due diligence is there to uncover the issues and reduce the possibility of very costly damages for unforeseen issues prior to acquiring or merging with, a target company. If the AI tool doesn’t have enough data in its bank of information to train the tool to locate relevant data, then lawyers are still going to produce a more accurate result. 

AI may also struggle to identify outliers. If there are definitions in documents that haven’t been correctly referenced or clause numbers that are incorrect, then the AI may not be able to detect all the relevant information. But even if a document is entirely accurate from a formatting and drafting perspective - it is rare, other than perhaps for non-disclosure agreements, for contracts to be entirely standardised. AI would therefore have to understand and be able to compare variations in language and style across contracts from different sources, to produce a comprehensive result based on the search parameters inputted. 

It's highly unlikely that the legal industry can fully standardise long form documents, as they are designed in their nature to be tailored to particular circumstances and achieve a negotiated position for each party. As a result of this, we may still require lawyers to analyse each of these contracts to ensure we can review all relevant data. 

Security risks

Another drawback, with potentially huge implications, is the security risk that using an AI tool represents. The nature of a M&A transaction means that very sensitive information is uploaded to a VDR and lays bare all corners of a target business in one digital location. As well as the required degree of security for the VDR - using an AI tool to analyse the data held in the VDR introduces another avenue for a cyber-attack that, in a worst case scenario, could lead to the leak of vital market data to a target company’s competitors. Although insurance and cyber security software can be arranged to deal with these attacks, it does not provide sufficient comfort to us or our clients if the highly sensitive data has already been extracted to a third party, for example the developer or operator of the AI.