A growing number of legal professionals now use artificial intelligence in their daily practice. This massive adoption conceals a more troubling phenomenon: the progressive erosion of verification skills, even for the simplest tasks.

The paradox is becoming clear. AI promises greater efficiency and productivity to law firms. But the first longitudinal studies reveal a perverse effect: professionals are losing their ability to detect errors in AI-assisted work, creating systemic risk for the quality of legal services.

The essentials

  • A significant proportion of legal professionals use AI, compared to a minority in 2022
  • Longitudinal studies show degraded verification skills even on simple tasks after 6 months of intensive AI use
  • Large law firms show higher adoption rates than solo practitioners
  • Automation primarily affects legal research, standard contract drafting, and document analysis

A lightning-fast transformation

AI adoption in the legal sector is following an exponential curve. A significant proportion of professionals now use AI tools, representing a considerable multiplication compared to 2022 figures.

This progression is explained by three converging factors. First, the technical maturity of specialized tools like Harvey AI, CoCounsel, or Spellbook, designed specifically for legal tasks. Second, competitive pressure from Legal Tech companies, which automate entire segments of standard legal advice. Finally, the immediate productivity gains observed on repetitive tasks such as case law research or drafting standard clauses.

Large firms are leading this transformation. Law firms with more than 100 lawyers show higher adoption rates than solo practitioners. This disparity reflects the financial and technical resources needed to deploy these solutions, but also the capacity to absorb the costs of training and team adaptation.

Usage concentrates on four priority areas: legal research, standard contract drafting, analysis of voluminous documents, and preparation of procedural documents. These time-consuming but relatively standardized activities are the natural targets for automation.

The silent erosion of critical skills

Behind this massive adoption lies a less visible but more troubling phenomenon: the progressive erosion of verification skills. A longitudinal study conducted over 18 months with several hundred AI-using lawyers reveals significant degradation in their ability to detect errors, even in simple tasks.

The study protocol was rigorous. Researchers compared the performance of two groups of lawyers with equivalent levels: one using AI extensively, the other maintaining traditional methods. After six months of intensive use, the first group showed a significant decline in its ability to identify factual errors in AI-generated contracts. After 18 months, this degradation was even more pronounced.

More troubling still, this erosion affects fundamental skills. Lawyers accustomed to AI struggle more to spot inconsistencies in legal references, date errors, or contradictions between clauses. Their critical vigilance dulls, replaced by excessive confidence in machine output.

This phenomenon is not specific to law. Generative AI adopted everywhere, productive nowhere documents similar effects in other professional sectors. Automation creates a cognitive dependency that weakens reflexes for human control.

Law firms are beginning to recognize the risk. Some are instituting mandatory double-verification protocols for all AI-assisted work. Others are organizing specific training on the biases and limitations of these tools. But these measures remain in the minority and are often applied inconsistently.

The productivity paradox: faster, but less safe

AI genuinely transforms law firm productivity, but not uniformly. Professionals using AI gain considerable time on administrative and research tasks. This freed-up time theoretically allows them to focus on higher-value activities: strategic advice, litigation, complex negotiation.

In practice, this redistribution of work time remains uneven. Senior lawyers effectively use this surplus to deepen their analysis or develop their client portfolio. But junior lawyers, traditionally trained through repetition of simple tasks, see their learning curve disrupted. They gain less practical experience in the fundamentals of the profession.

This situation creates a new type of professional stratification. On one side, established experts who use AI as a multiplier of their existing skills. On the other, a generation of young professionals who risk building their practice on fragile technological foundations.

The financial consequences are beginning to emerge. Several American law firms report an increase in client claims related to errors in AI-generated documents. Professional liability insurers are adapting their policies accordingly, with specific exclusions for damages related to unsupervised AI use.

The legal sector is discovering the limits of the “human-in-the-loop” model. Maintaining humans as final supervisors is only effective if those humans retain their error-detection capabilities. When this capacity erodes, the safety system crumbles.

Law firms experiment with new balances

Facing these challenges, several strategies are emerging in the profession. The most advanced law firms are developing hybrid approaches that seek to preserve human expertise while exploiting AI’s efficiency gains.

Allen & Overy, one of the first major firms to deploy AI extensively, established an “augmented apprenticeship” system. Young lawyers alternate between fully automated tasks and others deliberately maintained in traditional mode. The objective: preserve their analytical capacity while training them in modern tools.

Other firms are betting on technical specialization. They are training dedicated teams to audit AI outputs, creating a new profession at the intersection of law and technology. These “audit lawyers” develop specific expertise in detecting biases and errors in AI models.

Continuing education is becoming a strategic issue. Bar associations are beginning to integrate modules on AI risks into their mandatory training programs. Several bar schools are developing specialized certificates that attract a growing share of new registrants.

Some firms are experimenting with more radical approaches. They deliberately maintain certain non-automated practices to preserve their teams’ skills. This strategy, costly in the short term, could prove worthwhile if quality becomes an increasingly important differentiating factor.

Regulation still seeking its footing

Professional authorities are struggling to frame this rapid transformation. The challenge is twofold: encouraging innovation while protecting the quality of legal services and client interests.

The American Bar Association published AI usage guidelines in 2024, but without binding force. These recommendations emphasize the obligation of human supervision and the need to disclose AI use to clients. In practice, application remains highly heterogeneous across states and types of practice.

In Europe, the AI regulation adopted in 2024 classifies certain legal uses as “high risk,” imposing transparency and traceability requirements. But the practical modalities for implementation remain unclear, leaving law firms in regulatory uncertainty.

The French National Bar Council is working on a framework specific to legal professions. This text, expected in 2025, should define acceptable AI usage conditions and associated ethical obligations. The stakes are significant: reconciling technological innovation with the traditional requirements of the profession.

This regulatory uncertainty paradoxically hinders responsible AI adoption. Without a clear framework, some law firms prefer to wait, while others deploy these tools without sufficient safeguards.

Human expertise tested by automation

The irruption of AI in the legal sector reveals a challenge broader than simple technological adoption. It questions the very nature of professional expertise in the age of automation. How can critical skills be maintained when tools become more efficient than human analysis for certain tasks?

The answer does not lie in rejecting technology, but in redefining the role of the human expert. Tomorrow’s lawyer will need to master AI while retaining their capacity for critical judgment. This dual competence requires massive investments in training and a overhaul of professional learning methods.

Early experiments show that balance is possible. Law firms that invest in training their teams and maintain strict verification protocols manage to exploit AI’s gains without compromising quality. But this approach requires resources and a long-term vision that not all actors possess.

The stakes extend beyond the legal sector. Audit, engineering, medicine: all professions of responsibility face the same challenge. The ability to preserve human expertise while integrating AI will determine which sectors thrive in this transformation and which suffer a degradation of their added value.