In-Brief:
- The pace of development in artificial intelligence (“AI”) is reshaping corporate transaction trends, with increasing M&A activity targeting AI and technology-driven businesses.
- AI tools are already being utilized across various stages of corporate transactions, most notably in due diligence exercises, which can deliver greater speed and efficiency.
- While AI offers clear advantages in handling the ‘heavy lifting’ of corporate transactional work, its use will require legal professionals to be highly attentive to risks such as data confidentiality breaches.
In this article, we consider where AI currently is as well as the trends in corporate transactions, including AI companies and the AI-tools currently being utilised in corporate transactions. Given that the rate of development of AI is, arguably, the fastest-paced technological development in history, it may not be long before we need to return with an update and prediction on where AI is taking us.
The Current State of AI
AI has certainly been the buzzword for the past few years, and that doesn’t look set to change in the near future. The popularity with which ‘AI’ is being used to market products and services is only outstripped by the rate of development of new types of AI and applications making use of AI. If not already doing so, there seems little doubt that in the coming years everyone will interact more and more with AI.
In 2025, AI in current-use is popularly classified as one of either:
- Generative AI – which creates content, such as text, audio and music, images, and video; e.g. OpenAI’s ChatGPT and Dall-E, Anthropic’s Claude, Google’s Gemini, MidJourney, and StableAI’s Stable Diffusion; or
- Agentic AI – which processes data and makes, and dynamically adapts, its decisions to achieve a task; i.e. robots, automated driverless cars, ‘chat bots’ and customer service, and automated workflow management.
Text-creating Generative AI is often based on ‘Large Language Models’ (“LLMs”), which are machine-learning models that are trained to analyse and replicate patterns. LLMs are statistical mathematical functions that essentially predict and output the most probable text based on the input given. In order to do so, LLMs rely on analysis of large volumes of pre-existing text. The model is refined over time by more example text, fed back into the model in order to tweak the analysis and seek the most probable text, again and again, and again, to improve the accuracy of its output.
In legal practice, in many ways similar to the practice in the varied industries that commercial legal professionals serve, human habits will affect the rate of adoption of AI. Those that are willing to embrace AI tools will ultimately lead the conversion from traditional methods of delivering legal practice to clients.
Corporate Transactions and AI
In terms of corporate transactions and AI, there are both:
- trends in corporate transactions activity (mostly in M&A) involving AI companies; and
- AI being used by professional services in corporate transactions (again, mostly in M&A).
(i) Trends in corporate transactions activity involving AI companies
Over the next few years, as the pressure to adapt increases, we will undoubtedly see more buyers seeking to acquire AI capabilities through targeted acquisitions. In 2024, Goldman Sachs saw an increase to 12%, from a long-term average of 7%, in corporate transactions involving non-technology companies acquiring technology-focussed businesses. Whilst they attribute part of this to a broader rebalancing from ‘old economy’ sectors to technology, consumer, and healthcare sectors, they anticipate this rise to continue in 2025 as AI M&A activity gathers pace.
In terms of recent trends in AI M&A activity, market reports suggest that a significant number of transactions have related to acquisitions of companies behind AI ‘platform’ and ‘infrastructure’. The ‘infrastructure’ being data centres and the power required to supply AI servers and the ‘platform’ being the underlying AI models (some names of which will be familiar). In addition, the ‘application’ refers to the applications by which end-users engage with the AI platform. Goldman Sachs’ ‘2025 M&A Outlook Building Momentum on a Global Stage’ states that: “we view the AI era through an Infrastructure-Platform-Application framework, with most M&A activity to date concentrated at the Infrastructure and Platform layers.”
In general, Goldman Sachs term the growth in Agentic AI as “AI eats IT”. Put simply, if AI is able to execute tasks currently performed by humans there will be a decline in the value of IT companies that currently provide the IT infrastructure necessary for humans to conduct those tasks. Separately, as a result of the rise of Generative AI, they foresee the potential for decreased valuations of SaaS (Software as a Service) companies. Goldman Sachs’ ‘2025 M&A Outlook Building Momentum on a Global Stage’ states:
“Several potential “killer apps” have spurred clusters of targeted transactions, but agentic AI—in which models execute tasks currently performed by humans—has garnered palpable excitement and will catalyze dealmaking as "AI eats IT" across industries. Generative AI is expected to unlock efficiencies that will be deflationary and potentially disruptive for some SaaS companies—pushing down valuations and, in certain cases, driving them to go private. Conversely, software companies with entrenched customer relationships and proprietary datasets represent beachheads for AI transformation and have already produced solid M&A outcomes.”
(ii) AI being used in corporate transactions activity
In corporate transactions, AI is most frequently used in due diligence exercises, whereby LLMs can be used to ‘review’ and spot patterns and material provisions in contracts and other commercial documents. Harvey’s ‘Vault’ is one example of an application that aims to develop the standard virtual data room into one that can process and provide relevant insights from the stored documents.
In any due diligence exercise, the immediate benefit of using AI is the speed with which document reviews can be conducted. This is of particular importance given the tight timeframes in which a buyer’s counsel is often required to complete the due diligence exercise. This is of particular relevance to ‘full form’ due diligence exercises, whereby a buyer’s counsel is required to report full form reviews of all contracts and commercial documents provided in a data room and the use of AI can shorten the time period for carrying out such reviews. Prior to any formal due diligence, it is also common for a buyer to conduct initial due diligence on the target from publicly available information. In this case, Generative AI may be able to assist in preparing and summarising available public sources of information on a target. This will currently be of limited use in the UAE given the general lack of publicly available information on UAE-based companies. However, we expect that the preparation and summarisation of publicly available information will be substantially more prevalent in jurisdictions that do offer more publicly-available information on companies.
In addition to due diligence exercises, AI tools can be, and are currently, used effectively for the translation of foreign language documents. AI tools can also be effectively used to redact text, of a confidential nature, contained in certain documents.
However, an implicit risk in relying on AI tools to assist in corporate transactions is the possibility of AI tools reaching conclusions that a human would not arrive at from the same data. In a similar vein, the expertise that a human has in being able to determine the practical importance of, or weight to be placed on, various conclusions is not (currently) matched by AI tools. However, proponents of AI argue that ultimately, less time spent reviewing can equate to more time spent on analysis and (arguably) better conclusions.
As well as this, for more than two decades, buyers and sellers on corporate transactions have become accustomed to using virtual data rooms (“VDRs”) and the benefit of the strict confidentiality and accessibility protections that VDRs provide. It is a very different choice for a target to a transaction to volunteer to feed their most commercially sensitive data into an LLM, in which there is no real assurance as to what happens to that data after the transaction closes or how that data may have worked its way into the ‘learning experience’, and future output, of the LLM. Proponents of AI tools do appear to recognise this and, as an example, Intralinks’ ‘Dealcentre AI’ is a product that is based on its own LLM, and also provides assurances that one customer’s data is not used to re-train its own model.
In conclusion, we believe that AI tools can assist at this stage with what could be regarded as the ‘heavy lifting’ across many areas of legal practice, including due diligence for corporate transactions. However, as with many areas of legal practice, there is no black-and-white answer or one-size-fits-all approach, and our prediction is that in the future lawyers will need to be sharper than ever to direct and assess the weight of heavy lifting that AI tools are used to assist with.
Should you require any information on the above, please contact Patrick Tweedale (Partner), Christopher Fenn (Associate) or Alexander Wagg (Associate) at Hadef & Partners.