Does Your M&A Deal Need An AI Strategy?

By Hemant Krishna

M&A work is an acquired taste. Lawyers, bankers, and analysts — chasing the adrenaline of a closing — burn the midnight oil dissecting contracts, spreadsheets, and data rooms for that one buried risk or hidden value driver. Most of those who work on deals agree that M&A advisory is an endurance sport; it is about grinding it out. Perhaps that is why, over the past year, founders, students, and fellow lawyers have asked me the same question more times than I can count: what does artificial intelligence mean for transactional work — the bread and butter of bankers, funds, and corporate law firms? I must confess that my own perspective on that question has evolved through the experience of using AI across multiple deals.

Most AI vendors pitching to M&A practitioners offer the same promise: faster, cheaper, better delivery. But are they revealing the complete picture? Here is what my first-hand experience suggests.

What AI Already Does Well In Deals

AI shines where volume, speed, and pattern recognition are required. Sophisticated tools now widely available can classify contracts, extract clauses such as on change-of-control or assignment and compare them against precedents or a playbook in minutes — something humans take much longer to do. The benefit is practical: teams start from a near-complete set of facts instead of hunting for them. If parties have agreed to a bespoke arrangement with no ready precedent, AI can help craft a clause that avoids the tell-tale signs of a hurried drafter. Financial work benefits too, as AI ingests data-room exports, reconciles spreadsheets, and devises formulas with ease. None of this replaces judgment; it simply creates better inputs for human decision-making. 

AI-Integrated Workflows

In my experience, successful integration of AI into the M&A process is more plumbing than magic. Tools for analysis and comparison, de-duplication, OCR, and smart tagging make vast virtual data rooms more searchable and auditable. A source-referencing layer that anchors outputs in the underlying documents reduces the risk of hallucinations. Paired with clause libraries, precedent banks, and deal playbooks, AI helps create a workflow in which machines present facts and users can take decisions, drawing on knowledge, experience, and context. 

Guardrails And Best Practices That Actually Work

While much has been written about lawyers citing made-up cases in court thanks to AI hallucinating, there is lesser awareness about the guardrails needed for AI in the deal room. At its core, transactional work hinges on human judgment and value as well as risk. This makes human oversight non-negotiable. Client consent, privilege, and confidentiality also demand deliberate design, including segregated clean-team environments and vendor contracts that prohibit training on client data and commit to deletion on demand.

Even clients may not have the right to consent to processing a counterparty’s data for diligence. So, guardrails must be fine-tuned to the context of each deal, not set once and forgotten. AI outputs, of course, are only as reliable as their training and input data. Firms should think carefully about the training data a model relies on, and how much firm and client information users should input to improve results. Notwithstanding general guardrails, certain categories of highly sensitive data are best kept off the models.

Where Humans Must Lead

In the peculiar sport of deal-making, a lawyer is only as good as the bargain they can strike for the client. Negotiation is an art no model can own. AI can craft arguments and options, but it cannot weigh relationships, read counterparties, or trade a warranty for a price adjustment at the right moment. Strategy remains deeply human, because a business leader must decide whether a known risk is fatal, fixable, or a bargaining chip. Regulatory grey areas resist automation too; AI is unlikely to throw useful light on market practice or predict how a regulator will interpret a grey clause. In areas that require judgment and strategy, AI can support, but it cannot substitute for, human involvement. 

Regulation And Ethics In The Indian Context

Compliance and governance must be baked into technology, not bolted on after the fact. Personal data protection regimes emphasise lawful purpose, minimisation, and security; these should shape data-room design, processing geography, and vendor terms. Competition law and securities regulations often make clean-team protocols imperative and demand that AI insights from sensitive datasets do not leak to commercial teams. Bias and explainability matter too, because models trained on skewed precedents can overemphasise certain risks unless monitored and corrected. Above all, client trust depends on transparency about tool use. Therefore, as a standing rule, no client data should be exposed to public models without consent. 

Conclusion

AI in M&A has moved from pilot to practice, but it will not, at least in the near term, displace human expertise in the deal room. The right approach is “human-led, AI-enabled.” Let technology sweep the forest floor while professionals lead negotiation, strategy, and regulatory engagement. Make integration boringly robust with secure pipelines, source-referencing outputs, clean teams, and audit trails; all aligned with data, competition, and securities frameworks. Done right, the near-term prize is shorter diligence timelines and cleaner first drafts; the long-term value is better, more informed decision-making. 

If firms govern AI with the same care they bring to the rest of the deal, clients will get the speed they want and the judgment they need and deal-making might finally stop being an endurance sport.

The author is partner at Shardul Amarchand Mangaldas & Co. 

Disclaimer: The views expressed here are those of the author and do not necessarily represent the views of NDTV Profit or its editorial team.