The work of a lawyer is roughly divided into the stages of "understanding and defining the problem", then "research", and finally "drafting and advising", and the remuneration is calculated based on the time required for these tasks (in the case of incidents and accidents, there are cases of retainer fee and success fee, but they are omitted).
Among them, generative AI can achieve high efficiency and quality in "research" and "drafting and advice".
In English-speaking countries, services using generated AI are already provided for contract preparation and review work, research work such as court precedents, etc., and in Japan, LegalForce, an AI contract examination service, has started a service using ChatGPT.
In the next one to two years, it will be necessary to determine "what should be entrusted to AI" and "what lawyers should handle" and improve the accuracy of how to use the former (including improving prompt power).

The legal industry lacks datasets for generative AI to perform additional learning. In particular, the Japanese area has a low disclosure rate of court precedents, and there are few cases of publication of administrative documents in machine-readable formats.
The value of human resources who can collect and provide "perspectives" and "materials" that generative AI lacks, including documents used in court cases and undocumented information, will increase completely.
On the other hand, what is likely to become obsolete is the time charge model of law firms.
I believe that subscription-type services such as the legal consultation flat-rate plan and the contract review flat-rate plan will spread to the lawyer industry.
It would not be surprising if generative AI could be born that provides additional learning specialized in a specific legal field (e.g., fintech) and can provide the same level of advice as a first-class lawyer in that field (although in Japan, the way of providing services in relation to the lawyer law should be considered separately).