Harvey II adds memory & matter-based legal AI tools

Harvey has launched Harvey II, a new version of its legal AI platform that combines Memory, Spaces and agents.

The update focuses on tools designed to retain how individual lawyers work and organise work around specific matters or projects. The system is intended to start with a lawyer's existing preferences and case context already in place, rather than requiring repeated instructions for each new task.

Memory is the most prominent addition. It is designed to learn and retain an individual lawyer's drafting style, preferred structure, citation habits, level of detail and other working preferences over time.

Those preferences can carry across Harvey's platform, as well as Microsoft Word and Outlook, and can also inform tasks handled by software agents. Users can review what the system retains, change it or switch it off. Information stored in Memory is not used to train models.

Alongside Memory, Harvey II introduces Spaces built around a matter or project. These are intended to hold documents, tasks, permissions and work history in one place, giving agents and lawyers access to the same context as work progresses.

Tasks within a Space can be assigned to either a lawyer or an agent, with the next step moving to the relevant person or system. Permissions and ethical walls can sync from a firm's existing systems into a Space, while usage and cost remain attached to the same matter or project.

Agents can move between Spaces without carrying client data with them. That design reflects the legal sector's focus on confidentiality, client separation and internal access controls.

Built with firms

Harvey developed Memory over the past six months with input from law firms and in-house legal teams across the US, UK, Europe, the Middle East and Asia-Pacific. The design partner group included AmLaw and Magic Circle firms, as well as legal departments at Fortune 500 and FTSE companies.

That work focused on defining what the system should remember, how those memories should be applied and where the limits should be set. Memory was built to operate within the permissions, confidentiality requirements and ethical walls already used by legal organisations.

Mallesons was one of the firms involved in the design process and worked directly with Harvey on the Memory feature as part of the development programme.

"We actively partnered with Harvey on Memory because it enables our people to capture their preferences, standards and drafting approach once, then apply them consistently across matters and engagements. One of the most powerful unlocks is its ability to understand how we each work, not just respond to what we ask. Personalisation like this makes AI feel like a genuine extension of our lawyers, rather than just another tool. Capturing our preferences and writing style means we can also reflect our clients' needs and include them in the flow of how we work," said Michelle Mahoney, Chief Innovation Officer, Mallesons.

Three phases

Harvey said the Memory feature will roll out in three stages. The first, Personal Memory, is already in early access and focuses on the individual lawyer, including preferences around drafting, structure, tone, citations and ways of working.

The second stage will apply relevant memories across a shared matter, Space or vault. A later stage will allow firms and legal departments to bring broader ways of working into Harvey while maintaining boundaries between users, teams, matters and clients.

That phased approach points to a wider issue for legal AI providers: how to make systems more useful to individual professionals without eroding confidentiality rules or blurring internal information barriers. Harvey is positioning its product around that balance by linking personalisation to existing controls over access and client data.

The update also reflects a broader shift in legal technology away from one-off, prompt-based tools and towards systems structured around ongoing work. Instead of treating each request as a fresh interaction, Harvey II is organised around persistent matters, stored context and software agents that can act within those boundaries.

For law firms and legal departments, that could affect how AI is integrated into daily workflows, particularly in drafting, matter management and internal collaboration. The model also suggests legal AI companies are trying to move beyond document question-and-answer tools into software that takes a more active role in routine legal tasks.

The refreshed product experience in Harvey II brings matters, tasks, agents and lawyers together around the work itself, with usage and context tied to the relevant matter or project rather than rebuilt each time.