Experimentation with AI tools continues apace, whether ancillary to, or a critical component of a business’ core proposition. In the world of mergers and acquisitions, AI is increasingly being employed in the transactional due diligence (DD) process, becoming in some respects a staple of the work underpinning deal analysis. Advisers are deploying AI tools to extract and organise critical data from large volumes of information at scale and speed; work normally requiring significant time and human resource. With AI considered to be more efficient and cost effective when undertaking certain tasks, law firms and other advisers have invested heavily in their AI capabilities. Yet, for counterparties and the M&A insurers underwriting these transactions through warranty & indemnity (W&I)/representations & warranty (R&W) policies, clarity is needed around how AI tools are used, what their limitations are and how their outputs are validated.
Client and adviser demand for cost and efficiency savings will continue to drive AI innovation in the DD space, raising questions around how much diligence work AI will eventually handle and whether current systems are deployed in a transparent, controlled manner, and subject to expert human judgement. From an insurer’s perspective, the fundamental considerations fit into two categories: (i) how can we confidently support an AI-driven DD approach; and (ii) closely related, the extent to which AI is used and relied upon by target companies themselves.
The AI in DD advantage
AI in its current form is most effective when analysing large volumes of information – contracts, emails, reports etc. – to identify and/or summarise irregularities and patterns that manual reviews can miss. As well as specific DD tasks, AI tools are also being used effectively by commercial advisers to assess a target company's strategic positioning, product market fit, competitive dynamics and – increasingly – the potential for AI itself to disrupt the target's business model. Using AI to screen counterparties against sanctions lists, PEP databases and regulatory records is also increasingly common, work that has traditionally required significant manual effort by internal or external legal functions.
Overall, AI is allowing some advisers to conduct broader diligence in shorter timeframes, in some cases enabling more thorough exploration, with (in theory) no material increase in cost or headcount.
And the disadvantages…
Overreliance on flawed AI systems may lead to AI-generated mistakes or omissions, creating exposure which has the potential to become systemic. There are also ethical concerns, with autonomous tools requiring careful compliance oversight to prevent unethical or erroneous decision-making. AI-specific procedures, processes and guardrails are necessary to address inherent risks such as intellectual property breaches, algorithmic bias and ongoing compliance with evolving AI regulations. Serious issues may arise from the processing of sensitive or proprietary data by AI which, if robust safeguards are not in place, could be vulnerable to third-party misuse.
Potential disruption issues and ethical concerns must therefore be carefully evaluated and managed by advisers using AI in their diligence work. Also, given the speed with which AI has become embedded in many processes, it remains unclear in all cases whether an adviser’s professional indemnity insurance, which covers errors, omissions and/or negligence, will cover liabilities relating to work product generated by AI.
Conditions for insurer support
Early engagement with M&A insurers is a critical factor in their ability to confidently support a heavily AI-informed DD approach. They need to know how AI was used, to what extent and in what areas of review, all of which the scoping/basis of preparation sections of DD reports should expressly confirm. The details of the AI technology providers involved and the experience of the advisers in using the tools being deployed are further important factors, with assurances needed that AI outputs will, in any event, be verified by a qualified expert.
While proficient in processing and summarising large volumes of information, AI is less capable with tasks requiring judgement, interpretation or critical analysis. There is a broad consensus that AI is, at present, unsuitable for complex, technical areas of review where subjective interpretation is necessary, such as typically scoped tax diligence, financial diligence or specialist areas of legal due diligence. To support an AI-assisted DD process, M&A insurers therefore need advisers to provide a clear, coherent rationale for what it was used for and why. AI should be employed in a demonstrably logical and supplementary capacity – to assist a review, not to lead it.
With AI now a factor relevant in almost every deal, the extent to which AI is used and relied upon by the target company itself is a second key consideration for M&A underwriters. The main issues being how central it is to the business’ operations and what risks arise from that level of reliance. From an underwriting perspective, this requires a combination of legal and technical analysis, including scrutiny of data rights, intellectual property ownership, regulatory and data protection compliance and potential exposure to cyber risks. In addition, attention needs to be paid to how AI systems are developed and maintained, including any third-party licencing arrangements and the extent to which expertise is concentrated with key individuals and/or teams. AI is increasingly a significant factor in assessing both value and risk, with the depth of enquiry ultimately driven by how material AI is to a given target company and the extent to which the target actually owns any proprietary elements of the tools it uses.
In view of the fact that AI in its current guise remains a relatively nascent tool for many business sectors, as well as professional advisers, historic claims data directly related to its usage is limited. However, the potential for AI-related W&I/R&W claims is a live concern for insurers, with future losses potentially stemming from an AI ‘hallucination’ or miscalculation at a fundamental level, potentially giving false readings about critical aspects of a business’ health or data, upon which in turn wider (inaccurate) due diligence assumptions are based.
Generally, insurers are supportive of AI-assisted DD processes, provided advisers can demonstrate how the relevant tools were used and that the process was robustly governed. Caution remains around the extent to which this evolving technology is used, with AI expected to support rather than replace established diligence processes, and with a clear emphasis on transparency, defined scope, intent and expert human oversight. As AI tools develop and use becomes more embedded and standardised, these parameters are likely to evolve, informed partly by claims experience, as and when it emerges. Over time, there is huge potential for AI to provide broader and more comprehensive insights than current, manual-only DD processes can feasibly deliver. For now, confidence will depend on the same fundamentals that underpin any diligence exercise: clarity of approach, accountability for outcomes and the reliability of the work on which risk is ultimately assessed.
