Overview and Practical Takeaways
AI is accelerating M&A on multiple fronts. Buyers, for example, are increasingly using AI to source potential deals, filter and rank initial candidates, and then dive deeper into the most attractive targets.
But is AI’s impact on M&A automatically positive? The short answer is no. AI’s benefits and utility are undeniable. But AI can also introduce new legal and practical risks that must be carefully managed.
AI in M&A due diligence (DD) questions and answers is a prime example. Our key takeaways include:
- More is not better with DD questions. AI output can’t be allowed to overwhelm negotiations, distract from material value or risk, diminish human expertise, or induce a false sense of confidence in the quality of DD being conducted. Human judgment as to what really matters remains essential.
- Explainability is everything with DD responses. Over reliance by a seller on AI can raise red flags and undermine credibility. Anecdotal evidence of buyers walking away from deals because of AI-prepared DD responses is growing. Inaccurate DD responses can lead to liability in contract and/or in tort.
- Even with this complexity, the key principles M&A parties should follow remain fairly straightforward. Buyers should be able to justify why the DD question is being asked and the value or risk it addresses. Sellers should be able to explain the DD response given and the person, materials or other source that supports it.
Our detailed comments follow. For our thoughts on AI’s impact on RWI (representation and warranty insurance) in M&A, see here. For more Fasken M&A thought leadership, visit our Capital Markets and M&A insights hub and subscribe.
AI in M&A Due Diligence Questions: More is Not Better
There’s no doubt that AI adds significant value in generating DD questions and processing the seller’s corresponding responses. AI can quickly produce a first draft of DD questions based on precedent and then begin to customize them for the target, including based on any confidential information memorandum (CIM) and/or management presentations. AI can compare DD questions across workstreams to identify gaps, overlap or inconsistent terminology. It can monitor seller responses, suggest follow-up questions, and track outstanding or unresolved points. It can also flag problematic issues, e.g., potential contradictions among seller responses or between seller responses and target documents.
But, as with many other AI applications, AI is merely a tool and not a solution. Its ultimate value depends on the skill of the hand guiding it.
On a practical level, more is not better in the context of DD questions. The focus must always be on quality and not quantity. Excessive DD questions can frustrate sellers and damage negotiation dynamics. They can also overwhelm target management, unreasonably increase costs, and unnecessarily delay the deal process. The buyer’s aim should be tightly knit and highly contextualized DD questions that drive directly at value or risk in search of meaningful insight.
The pursuit of this goal necessarily requires situational awareness and human judgment. DD questions should be tailored to the target’s industry, the transaction’s size and the deal’s timeline, among other things. Effective diligence is also judgment-based and iterative. It should be informed by contextual cues beyond the substance of DD questions and responses (e.g., does the seller appear to be stalling on the point?). AI may miss what seemingly benign answers may warrant deeper investigation. Similarly, AI may overstate the significance of an issue. Matters best left for human analysis should include the classification of DD issues, such as for priority, materiality and potential purchase agreement implications. Ultimately, only deal counsel can give legal advice against the backdrop of what really matters to a particular client in the particular transaction.
Applying AI in buy-side diligence also requires careful management of the various risks that accompany AI use. Only appropriate AI tools should be used, with safeguards in place to ensure that sensitive target information, transaction strategy and legal analysis aren’t compromised. The quality of AI output also fundamentally depends on the quality of the inputs. Part of being good deal counsel in the AI age is knowing good prompts from bad and how to get the most out of AI in the acquisition at hand. AI’s potential “agreeableness” (i.e., sycophantic tendencies) must also be carefully managed. A risk here is the potential for AI to reinforce rather than challenge the buyer’s deal thesis.
Overall, and unsurprisingly, the key is for buyers not to become over reliant on AI diligence tools. AI’s speed and volume can’t be allowed to derail negotiations, distract from material value or risk, diminish the importance of human expertise or experience, or induce a misplaced sense of confidence in what is in truth substandard diligence.
AI in M&A Due Diligence Responses: Explainability is Everything
One might be tempted to view AI use in DD responses in the same light as AI use in DD questions, as if they were two sides of the same coin. This would be wrong. While certain of the issues raised overlap, there are important differences. Critical here is that poorly drafted DD responses carry potential risks and liabilities that poorly drafted DD questions do not.
But let’s begin with the value that AI, aptly used, can bring to DD responses. As with DD questions, AI can quickly prepare first drafts of DD responses from the target’s internal records. AI can help coordinate DD responses from across different working groups. AI can identify potentially inaccurate responses or potentially conflicting issues across different proposed responses. It can highlight when a proposed response doesn’t properly answer a question or may lack factual support. It can flag where a proposed response may go beyond the question asked, e.g., by an unnecessarily overbroad statement. AI can also keep tally of answered and outstanding questions and advise as to potentially necessary updates as time moves forward.
Against these benefits, the different risks that AI use entails must be weighed. Perhaps most well-known is the risk of AI hallucinations. Similarly, AI can infer facts, seek to fill what it sees as gaps, merge what should be properly kept as distinct concepts or topics, and/or make assertions with more confidence than the facts warrant. AI generated DD responses can contradict discussions that have taken place outside the information the AI has access to. The AI’s analysis can also smooth over potentially problematic issues within the target’s business that instead deserve the attention of target management.
Once again, mitigating these risks requires appropriate human involvement in AI’s role in the DD response process. This is also necessary to ensure all prudent security protocols are followed, and that sensitive company, employee, customer, or transaction information is not inadvertently inputted into AI outside appropriate guardrails. The potential for “shadow AI” use by target or seller personnel in connection with the DD response process is a particularly pernicious risk. Human oversight is also important to guard against the inadvertent disclosure of legally privileged communications.
In terms of deal dynamics, an apparent over reliance by a seller on AI in preparing DD responses can quickly raise red flags and undermine credibility. This can adversely impact pricing and/or the buyer’s desire to proceed, among other things. Concerns of this nature can stem from the substance of the seller’s DD responses or in the M&A parties’ related discussions, e.g., in the buyer’s discussions with target management or in follow-up DD calls. Buyers will expect (1) a seller’s DD responses to reflect management’s actual knowledge, and (2) sellers to be able to explain and defend the responses they’ve given.
Anecdotal evidence of buyers walking away from deals because of apparent seller over reliance on AI in the DD process is growing. Two examples recently cited by Pitchbook are illustrative. In the first, an investor withdrew from a live sales process “after receiving diligence responses from a management team that were clearly AI-written and unrefined.” In the second, the deal fell apart after a “founder was discovered to have used ChatGPT to answer fundamental questions such as how his business reached its current growth rate.”
In terms of potential adverse legal consequences, inaccurate or misleading DD responses can give rise to potential liability in contract (i.e., under the purchase agreement) and/or in tort (e.g., for misrepresentation or even fraud). It should also be appreciated that, in RWI deals, the seller’s DD responses will, together with the other diligence documents, inform the scope of the policy and its exclusions (and will be relied on in making associated no claims declarations). Lastly, and once again, AI’s potential “agreeableness” (i.e., sycophantic tendencies) must be appreciated and carefully guarded against. In the context of DD responses, a risk is that AI could impulsively seek to ignore or downplay problems rather than detail their true dimensions. Pitchbook has put it this way: “AI is built to agree”.
Concluding Comments: Understand What You’re Asking/Answering and Why
DD questions and DD responses are both fundamental to M&A transactions, but they raise several different practical and legal considerations.
DD questions are principally an investigative tool and from a legal perspective are, in and of themselves, essentially inert. Their main function is to help the buyer better understand the target and the potential risks associated with its acquisition. Where a buyer’s DD questions are subpar, the principal consequence is an information deficit such that the buyer can’t negotiate as effectively and may assume otherwise avoidable risk.
DD responses raise more complex issues. Yes, as with poorly managed DD questions, poorly managed DD responses can adversely impact deal dynamics and deal momentum. But poorly prepared DD responses can also create seller liability to the buyer where none existed before. Depending on the circumstances, this can lead to damages in favour of the buyer and/or to a buyer’s right to avoid closing the transaction. Moreover, should the misrepresentation have induced the deal or should it amount to fraud, the buyer’s rights can extend to a right to rescind a transaction post-closing and/or a claim against the seller individual(s) responsible for the false statement in their personal capacity.
Fortunately, for both buyers and sellers, the guiding lights to follow remain relatively straightforward. Harness the utility of AI in preparing DD questions and responses, but don’t let AI use detract from your ability to clearly explain the most basic issue informing the exercise at hand. For buyers, this is the reason why you’re asking the DD question and the value or risk it addresses. For sellers, this is the person(s), materials or other sources that substantiate the DD response and how.
AI is supercharging DD in M&A. But more capability only increases the importance of adept control. Similarly, while AI can accelerate both sides of the DD equation, human judgment must still drive the deal.
