AI in Deal Teams: Hired for Speed, Applied for Strategy
The increasing reliance on AI raises a question worth pausing over: Does speed always translate into clarity, or can it sometimes introduce a different kind of uncertainty?
Artificial Intelligence has quietly taken a seat in the deal room. In many transaction teams today, AI tools function almost like an additional junior associate-fast, tireless, and capable of processing volumes of documents that would otherwise require days of human effort.
The advantages are obvious. In large due diligence exercises, AI can scan hundreds of agreements, extract key clauses, categorize risks, and produce structured summaries within minutes. Work that once required long nights of document review can now be completed in a fraction of the time. For deal teams operating under tight timelines, that efficiency is difficult to ignore.
Yet the increasing reliance on AI raises a question worth pausing over: Does speed always translate into clarity, or can it sometimes introduce a different kind of uncertainty?
A useful way to examine this question is through the different layers of transactional work where AI now operates.
Efficiency in the Mechanics of Diligence: Consider a typical diligence exercise in a portfolio acquisition: whether in real estate, infrastructure assets, or an operating business. Hundreds of documents must be reviewed: commercial contracts, leases, financing arrangements, regulatory approvals, and vendor agreements.
AI tools perform well in precisely this environment.
They can identify patterns across large sets of documents: the termination rights, change-of-control provisions, exclusivity clauses, indemnities; and present the findings in a structured format. Instead of manually locating clauses across hundreds of agreements, lawyers receive an organized map of where potential issues lie.
For instance, in a portfolio of several hundred tenant leases, an AI review may quickly indicate that most agreements follow a similar termination framework, say, six months’ notice with a defined penalty. From a pattern-recognition perspective, this provides immediate comfort around standardization.
The challenge arises in the minority of agreements that depart from that standard.
AI may categorize such deviations as routine drafting variations. A lawyer reviewing the same provisions may see something quite different.
Like a junior associate, AI can identify issues, summarize documents, and flag deviations. Deciding which of those issues truly matter remains a human responsibility
Reading Clauses vs. Reading Consequences: Take a situation where a handful of leases allow termination upon “material redevelopment of the premises.” At first glance, the wording may appear similar to other termination clauses and may therefore be grouped by the system as standard.
But a closer human review may reveal that redevelopment is defined broadly enough to include structural changes triggered by zoning adjustments or regulatory upgrades. In a city where planning regulations evolve frequently, such wording could permit tenants to exit long-term leases much earlier than expected.
For an acquirer evaluating rental stability and long-term asset valuation, that distinction is not merely academic.
In simple terms, AI reads clauses; lawyers read consequences.
A similar issue often emerges in operational contracts. In the acquisition of a manufacturing business, an automated review of supplier agreements may indicate that indemnity structures are broadly consistent across vendors. Yet a detailed human review may uncover cross-default triggers tied to financing arrangements or change-of-control events.
These clauses are often embedded within otherwise standard provisions. But their commercial implications can extend well beyond the contract in which they appear. A financing restructure following an acquisition could, for instance, inadvertently activate termination rights across key supply relationships.
It is precisely these second-order implications that require judgment rather than pattern recognition.
The Problem of Creative Drafting: Transactional practice has always relied heavily on what might loosely be described as creative drafting.
Lawyers frequently structure obligations through layered provisos, indirect references, and negotiated carve-outs. A clause may appear entirely conventional until a single phrase subtly shifts the risk allocation.
Consider a limitation-of-liability clause that caps exposure except in cases of fraud, wilful misconduct, or breach of confidentiality. At first glance, this resembles a standard carve-out.
But if confidentiality is defined broadly elsewhere in the agreement, perhaps including operational data, commercial information, or internal strategy, the carve-out may effectively expose the party to uncapped liability across a wide range of circumstances.
Experienced lawyers tend to pause at such language almost instinctively.
Algorithms, at least in their current form, often treat these variations as part of the drafting landscape rather than as potential risk reallocations.
The same phenomenon appears in joint venture agreements. Governance rights are sometimes structured in ways that indirectly grant minority investors effective veto power—through quorum requirements, reserved matters, or carefully drafted approval mechanisms.
These arrangements rarely announce themselves clearly within the text. Their implications emerge only when the provisions are read together.
Context Beyond the Contract: An even more complicated layer lies beyond the contractual language itself.
Many businesses, particularly promoter-driven or relationship-based enterprises, operate within commercial frameworks that extend beyond the four corners of the agreement. Contracts record the legal arrangement, but day-to-day practice may evolve through informal understandings or long-standing commercial habits.
For example, a distribution agreement may formally restrict sub-distribution without prior written consent. Yet in practice the distributor may have historically appointed local partners with tacit approval from management.
An AI tool reviewing the contract will correctly identify the restriction. What it may not detect is that the restriction has never been strictly enforced.
Deal lawyers often develop this contextual understanding through conversations with management teams, operational staff, and industry participants. Those inputs then inform the overall risk assessment.
Algorithms trained on structured contractual text are not particularly well positioned to capture that layer of commercial reality.
Regulatory Interpretation: A related issue arises in regulatory analysis.
Regulators rarely interpret transactions purely through the literal language of documentation. They tend to look at substance, intent, and the overall commercial structure.
In foreign investment, competition law, or sectoral regulatory regimes, a structure may technically satisfy formal thresholds while still raising questions about effective control or indirect influence.
Similarly, in infrastructure or real estate transactions, documentation may comply with statutory requirements but still create arrangements that regulators view as inconsistent with policy intent.
AI tools can verify that contractual language aligns with prescribed regulatory parameters. But assessing how regulators might interpret the broader commercial structure remains a distinctly human exercise.
The Fastest Associate in the Room: None of this diminishes the value of AI. Its ability to reduce repetitive effort and surface patterns has already made it an indispensable part of modern deal practice.
If anything, the most accurate comparison is that AI resembles a particularly capable first-year associate, exceptionally efficient, able to produce impressive volumes of work, yet still requiring supervision and judgment from more experienced lawyers.
Like a junior associate, AI can identify issues, summarize documents, and flag deviations. Deciding which of those issues truly matter remains a human responsibility.
Over time, this may reshape the structure of deal teams. Tasks that once required large associate groups may increasingly be handled through AI-assisted workflows, allowing lawyers to focus more on negotiation strategy and risk assessment.
In that sense, AI may not reduce the importance of lawyers. It may simply move the profession further up the value chain.
Learning Machines, Responsible Lawyers: That balance will likely evolve quickly. AI systems are improving rapidly, and their role in legal work will continue to expand.
Which makes one final point worth noting.
AI learns from the material it is trained on. The agreements we draft, the negotiations we conduct, and the reasoning we embed in today’s documents will shape the datasets on which tomorrow’s systems are built.
If AI is learning from lawyers, the profession must remain deliberate about what it teaches; and cautious about what it allows machines to decide.
Because while AI may be the fastest associate in the deal room, the responsibility for judgment will continue to belong to the lawyers sitting beside it.
Disclaimer – The views expressed in this article are the personal views of the author and are purely informative in nature.