Back
Chander Uday Singh Warns AI Hallucinations Complicate Fee Rationalisation for Law Firms
The integration of generative artificial intelligence into legal research has sparked a fierce debate over how law firms should bill clients when technology dramatically reduces the time needed for routine tasks. At the centre of this conundrum is a fundamental tension: clients, seeing faster work, demand lower fees, while lawyers worry that the very speed of AI masks serious risks—including the . has sounded a cautionary note, warning that AI’s tendency to “hallucinate” authority poses a hidden cost that a simple percentage discount cannot address.
The issue came to the fore during a recent panel discussion, where a participant referred to as argued that clients are entitled to share in the efficiency gains delivered by AI. “’s concern is the saving created by faster research,” the source notes. The instinct behind cutting the bill because the work got faster is understandable, yet it is also, as the panel observed, “the wrong place to start.” A blanket discount treats every hour of AI-assisted work as equally cheapened, when in truth the quality and reliability of that work vary enormously.
The AI Efficiency Paradox
For decades, law firms have billed by the hour, a model that rewards time spent regardless of outcome. AI tools now enable associates to scan thousands of precedents in minutes, summarise complex statutes, and draft initial memoranda with unprecedented speed. From a client’s perspective, this should translate directly into lower invoices. Why pay for ten hours of research when a machine can do it in ten minutes? That logic, while superficially compelling, ignores a critical nuance: the output of generative AI is not uniformly trustworthy.
The source quotes as raising the alarm that “generative tools occasionally invent case law that was never delivered and shape answers around what the questioner wants to hear rather than what the law actually holds.” This phenomenon, widely referred to as , is particularly dangerous for junior lawyers. As Singh pointed out, juniors are “least equipped to catch” such errors. An associate who relies on a chatbot for case citations may inadvertently cite a non-existent judgment, potentially exposing the firm to sanctions, malpractice claims, or a damaged reputation.
When the Machine Gets It Wrong
The legal profession has already seen embarrassing episodes where AI-generated briefs included fictitious cases. In one widely reported U.S. instance, a lawyer used ChatGPT to prepare a motion and ended up citing several fabricated decisions. The court imposed sanctions, and the lawyer’s credibility suffered lasting harm. While such incidents remain rare, they underscore a systemic risk: the more law firms adopt AI for cost-cutting, the greater the temptation to skip the rigorous verification that human researchers traditionally perform.
Singh’s warning is therefore not merely theoretical. It points to a structural problem in the way law firms currently deploy AI. Most tools are trained on vast corpora of legal texts, but they lack the ability to distinguish between a and a persuasive , let alone a completely invented case. Moreover, AI models are designed to produce plausible-sounding answers; they do not “know” when they are uncertain. This means that every AI-generated output must be treated with suspicion, a reality that imposes a hidden cost: the time and expertise required to double-check the machine’s work.
A Client’s Perspective
’s demand for a discount is equally valid from a client’s standpoint. If a task that previously took twenty hours now takes two, why should the client pay anywhere near the old rate? The answer, according to the panel, lies in the value of the work, not the time it consumes. But value is notoriously difficult to quantify in legal services. A two-hour research session that uncovers a decisive precedent may be worth far more than twenty hours of mediocre work. Conversely, a quick AI-generated memo that misses a critical point could cost the client far more than any fee savings.
The real answer to ’s demand, the source suggests, “does not sit in a percentage knocked off in an invoice. It sits inside every firm’s own time-sheets.” In other words, law firms must first understand exactly how AI is affecting their workflow before they can design fair pricing models. Some tasks may be genuinely accelerated with minimal risk—for example, locating a known statute or checking a citation. Others, such as constructing a novel legal argument, may still require substantial human oversight, negating much of the supposed efficiency gain.
Rethinking the
The debate over AI and fees is forcing law firms to reconsider the itself. Alternative fee arrangements, such as , , or , are gaining traction as a way to align incentives. Under a fixed-fee structure, the firm bears the risk of inefficiency but also reaps the reward of AI-driven savings. That could eliminate the client’s demand for a discount, since the price is agreed in advance. However, require accurate cost prediction, which AI’s variability makes difficult.
Another approach is to charge based on the value delivered, rather than the time taken. This would require firms to articulate the specific benefit of their work—a challenging task when outcomes are uncertain. Yet it may be the only way to fairly account for the dual nature of AI: it can save time on routine tasks while increasing the risk of error on complex ones.
Impact on Legal Practice
The AI fee conundrum has practical implications for every law firm. First, it underscores the need for robust training and supervision of junior lawyers. As Singh noted, the least experienced members of a team are most vulnerable to AI’s pitfalls. Firms must invest in quality assurance protocols, such as requiring all AI-generated citations to be independently verified against primary sources.
Second, it raises ethical questions under professional conduct rules. Lawyers have a duty to provide , which includes ensuring the accuracy of legal research. Relying on unchecked AI output could violate that duty. Moreover, billing for time that is inflated by inefficient manual research, when a faster AI alternative exists, may be seen as unreasonable. The and other bodies have begun issuing guidance on AI use, but the law is still evolving.
Third, the debate highlights a growing divide between large firms that can afford sophisticated AI tools and smaller practices that may struggle to keep up. Clients may pressure all firms to adopt AI and reduce fees, but smaller firms might lack the resources to implement proper verification systems. This could lead to a two-tier market where only well-funded firms can safely use AI at scale.
Conclusion
The AI conundrum for law firms and their clients is not about whether to adopt technology, but how to price its use fairly and safely. As the panel discussion made clear, neither a blanket discount nor a blanket rejection of AI serves the interests of justice. ’s warning about reminds the profession that speed without accuracy is worthless. ’s demand for a share of the savings reminds firms that clients will not indefinitely pay for inefficiency.
The path forward lies in transparency, careful measurement, and a willingness to move beyond the . Law firms that track exactly how AI affects each phase of a matter—and communicate those insights to clients—will be best positioned to craft pricing that reflects genuine value. Until then, every invoice that includes AI-assisted work will be a lightning rod for debate. And every lawyer who signs off on that work must remember that the machine’s confidence is no substitute for their own .
