|
|
AI can already do much of the analysis in law and audit. It cannot sign the opinion, and that, not how hard the work is, decides which jobs stay human. |
|
One Number
The share of professional tasks, across 44 occupations, on which OpenAI’s GPT-5.4 matched or beat industry experts. OpenAI, GDPval (OpenAI’s own benchmark), March 2026 Its predecessor scored 71% three months earlier. The analysis is catching up with the experts. The signature is not. |
|
|
One Argument The usual test for an AI-proof job is how hard the work is to automate. A better test is who goes to court when it goes wrong.The question is no longer whether AI can do professional work. It is how often professionals already use it. Thomson Reuters surveyed 1,816 lawyers, accountants, auditors, and tax and compliance professionals in 62 countries this spring. 74% use AI tools several times a week, and 44% do that several times a day. The labour market has followed. Stanford’s Erik Brynjolfsson and colleagues used payroll data through June 2026 and found that employment for 22-25-year-olds in the most AI-exposed jobs is 19% below where it would otherwise be. This number has increased from 15% a year earlier. Experienced workers, however, show no such gap. Firms have cut entry-level hiring, not the people who sign. Liability, meanwhile, has not moved. Damien Charlotin’s database at HEC Paris now lists 2,184 court decisions in which someone filed material AI had invented, 864 of them by lawyers. The sanction here lands on whoever signed. In October 2025 Deloitte agreed to refund part of its fee to Australia’s employment department after a report it delivered cited non-existent research and a fabricated quote from a court judgment, according to Fortune. In 2025 the UK Solicitors Regulation Authority authorised Garfield.Law as the first law firm to deliver legal services entirely through AI. But it still required named solicitors to remain “ultimately responsible for all system outputs”, according to an analysis by Dechert, another law firm. Even the AI law firm has a person on the line. Liability is a cost. And firms can carry it. Deloitte paid the refund, not a named partner. Since May 2025, Lloyd's insurers have underwritten cover from the start-up Armilla against AI tools getting things wrong, the Financial Times reported. If insurance can absorb the risk, the signature becomes a formality. However, the insurance market is moving the opposite way. W. R. Berkley, a US insurer that wrote $12.7 billion in premiums last year, has drafted an “absolute” AI exclusion for directors’ and professional-liability cover. Together with Great American and Chubb, it has sought US regulators’ approval to carve AI risks out of corporate policies, with one underwriter calling model outputs “too much of a black box”, the same paper reported. When insurers step back, the risk falls on whoever signed.
|
|
One Position Stop judging whether a business or a career is AI-proof by how hard the work is. When you review a professional-services supplier, the useful question is not “how much of this can AI do?” It is “who signs, and what are we paying that signature for?” I would change my mind the day a regulator lets an AI firm operate with no named professional answerable for the output. If your signature were the only part of your job AI couldn’t do, what would you charge for it? And what would anyone pay? |
|
If this was forwarded to you and you'd like to receive Corporate Financier's Notes every Thursday, you can subscribe here. Reading on your phone? Tap the three dots or overflow menu in the top corner of your email app and select View in browser for the best experience.
|