Goldman Sachs: Only 1.7% Legal Jobs At Immediate AI Risk
A new analysis from Goldman Sachs suggests that only about 1.7% of legal jobs are currently at risk due to the advances in AI-driven automation. This figure marks a significant departure from the investment bank’s earlier projections, which, just two years prior, estimated that up to 44% of lawyer roles could be automated. Such a dramatic revision prompts a closer look at the immediate impact, or rather the surprising lack thereof, of generative artificial intelligence on professional employment.
While the legal sector appears largely shielded for now, more immediate data indicates that generative AI is indeed influencing hiring trends in other areas. Goldman Sachs, after reviewing official US employment figures, noted a reduction in hiring, particularly for junior computer programming positions. Graphic designers and telephone call centre workers have also experienced the effects of increased automation. However, for most professional roles, including legal, the immediate threat of automation by AI generally hovers between 0.5% and 2.5%, with tax advisers facing the highest exposure and finance experts the least. Even at its peak, the current level of impact remains remarkably low.
Furthermore, these potential job losses in professional services may well be absorbed and even surpassed by new employment demands as the global economy continues to expand and evolve. The underlying reason for this resilience, particularly in fields like legal advice, appears to be their heavy reliance on high levels of accurate human judgment. Roles where the totality of tasks can be completely handled by AI face genuine risk, impacting the job market until the economy adapts. Conversely, if a role retains even a small component requiring nuanced human judgment, it becomes considerably more difficult to automate entirely, even if the professional heavily leverages AI tools.
Consider the example of a junior engineer whose coding tasks could be 99.999% replaced by generative AI-produced code; such a role faces immediate obsolescence. There is no protective barrier. However, if legal AI tools are deemed only 90% as effective as a human lawyer’s work, that lawyer’s position remains secure. This is because the “last mile” of capability—achieving the remaining accuracy, completeness of review, and subtle legal understanding required to reach near-perfection—is incredibly challenging for AI. Unless an employer is willing to accept substantial risk, they will continue to rely on human expertise, especially from junior lawyers.
Consequently, the estimated 1.7% of lawyers deemed replaceable by AI today are primarily those engaged in low-level work, where their professional input or judgment is minimal. While a 1.7% loss of all US lawyers would still amount to approximately 22,000 jobs, it is possible that market expansion, a phenomenon akin to Jevon’s Paradox where increased efficiency leads to greater demand, could create new opportunities within the legal sector.
The long-term outlook remains far more uncertain and hinges significantly on future advancements in AI accuracy. Should AI tools achieve 99% accuracy and completeness across vast swathes of legal work, the landscape would fundamentally transform. However, that level of sophistication is not yet a reality. For now, the news is largely positive for legal professionals, but it is crucial not to assume that AI accuracy levels will remain static; legal AI tools are continuously improving.