// wiki · analysis · September 2026
The future of the legal profession
Draft article. Analytical material: forecasts are marked as forecasts, facts come with sources.
A draft for discussion. The article is a view of the shifting of roles, not an employment forecast: research and practice data show a direction, not guarantees.
The short answer
The profession is not disappearing — it is being redistributed. The model takes drafts, search and digests; the lawyer's value shifts to what the model does not do: setting the task, verifying the result, responsibility and judgement. The 2026 court practice fixed exactly this division: the lawyer who signs answers for the AI text (court practice review), and working with models has become a separate skill.
What the data says
A survey of 121 heads of legal functions (Deloitte UK and RSGI, June 2026) records growing adoption and budgets with a lag in governance maturity: tools are bought faster than rules for checking them are built — a data review in the knowledge base. The court system answered symmetrically: the Plenum of the Supreme Court obliged disclosure of AI use, and the fine for non-existent cases “selected” by a service showed the price of a skipped check.
What compresses
- primary search and selection of norms “from scratch” — the model assembles a draft picture in minutes;
- standard forms and first drafts of contracts;
- digests of multi-page documents and version comparisons;
- template letters and memos.
This is not the disappearance of work — it is the disappearance of its bottom layer. A lawyer who only moved texts around loses the role; a lawyer who decided things gets an accelerator.
What grows
- Verification: the ability to find hallucinations becomes a basic skill — verification protocols like the academic one enter the work standard;
- Task setting: the quality of the result is determined by the question; formulating requests is legal technique, not programming;
- Managing the instruments: AI policy, data contours, choosing services, working with assistants — new duties of the legal function;
- Product and process: legal service design, process automation, work metrics;
- AI law: companies need lawyers who understand both regulation and technology (the regulation map, the nature of models).
What does not change
Responsibility stays human — the courts have already fixed this directly. Not replaced are judgment in uncertainty, negotiation, the client's trust: the model averages the known, the lawyer decides in the specific case. Ethics and conflicts of interest do not go away — they get new occasions.
What to do right now
- master the protocol for verifying AI answers and require it of the team and contractors;
- adopt an AI policy: instruments, data, disclosure (the measures);
- rethink the training of junior lawyers: routine is gone — the price of mentorship has grown;
- start keeping a library of the team's verified instructions and prompts.
Team training is the direction of diff.academy; the rules of AI use in a company — ongoing support.