AI in law firms: from rejection to real implementation — three phases according to Olivier Shadyuto

Parisian consultant Olivier Chaduteau, founder of an AI-native consulting firm, identified three distinct stages in the legal sector's relationship with artificial intelligence. His diagnosis is harsh: most law firms are only now exiting the second phase — and finally doing so seriously.

Three stages: from denial to action

The first stage was predictable: lawyers rejected AI as inapplicable to their profession. Arguments sounded roughly like this: "Law is too complex and nuanced a field," "the client pays for our unique expertise, not a machine," "the court won't accept a document written by an algorithm." This phase lasted longer than in most other professions — partly due to the profession's high self-perception, partly due to real regulatory risks.

The second stage proved more candid about the nature of the industry. Organizations massively purchased licenses for LLMs — but not for actual work, rather for demonstration. At conferences, beautiful slides about "AI transformation" appeared, press releases about partnerships with OpenAI or Microsoft were released. Inside firms, a few enthusiasts experimented with ChatGPT, the rest worked as before. Licenses were paid from marketing budgets — to signal partners and clients, not for results.

The third, mature phase is beginning now. Its markers are concrete ROI metrics, specialized tools for specific tasks, and mounting pressure from clients who are already aware of AI capabilities.

Where AI is already really working

Real applications in the legal sector today are concentrated in several clearly defined areas:

• Legal research — precedent search, analysis of case law (Harvey.ai, Casetext, LexisNexis+)
• Contract review — automatic identification of risky clauses, comparison with templates
• Due diligence in M&A — processing thousands of pages of documents in hours, not weeks
• Drafting — initial versions of standard agreements, NDAs, employment contracts
• Billing and time tracking — classification of working time by case without manual entry

A key shift in economics: before, partners viewed AI as a threat to the billing model — fewer hours on a task meant a lower client bill. Now the logic has reversed. Freed-up time is redistributed to complex strategic work, while routine tasks are repriced at competitive fixed rates.

Clients became the catalyst

Large corporate clients — banks, investment funds, industrial holdings — now directly ask their legal advisors: "What AI tools do you use on our matters?" This pressure from below proves more effective than any internal initiatives. Compliance departments and D&O insurers also begin imposing new requirements for work speed and documentation.

In parallel, there is consolidation in the tools market. Instead of dozens of disconnected pilots, large firms select 2-3 specialized Legal AI platforms and integrate them into real workflows. The era of experiments is ending — the era of operational scaling is beginning.

What this means

The legal profession has come a long way: from categorical denial through window-dressing purchases to real automation. For practicing lawyers, this is not a threat of disappearance, but a shift in priorities: less routine research and groundwork, more strategic expertise. For Legal AI companies, the moment of truth has arrived — the market is ripe, clients are ready to pay, and the question is no longer "is AI needed in law," but "which competitor will win."