Today we are giving a name to something Donna has been doing for a year: working as an agent. By that we mean a system that takes a whole piece of work, plans it, carries it out with the matter’s real documents, checks its own output and hands the result to the professional who signs it.
The word agent is used widely in the industry right now, and often loosely. So before we describe what has shipped, we want to be precise about what we mean. The difference between a model, a model system and an agent decides how much of a piece of legal work can be handed over, so it is worth setting out.
How We Define AI Systems at Donna
The building block is a model: a frontier large language model. We use a combination of models from Anthropic and OpenAI, chosen per task. Given a single prompt, a model produces a single response. It is highly capable, but it is stateless: it knows nothing about your matter, keeps no memory between requests and cannot take action in any system.
Connect a model to task-specific tools and knowledge, and you get a model system. When you ask Donna a question about a contract, a model system retrieves the right passages from the matter’s documents, reads them and answers with citations. A model system runs a fixed pipeline of model calls and handoffs from start to finish, the same way every time.
An agent works differently. You give it an objective rather than an instruction, such as preparing the disclosure summary, chasing the outstanding conditions or building the settlement timeline. The agent decides what steps that takes, performs them with tools, inspects its own output and loops back when the work does not hold up. In Donna, one more thing is part of the definition itself: the loop ends at a person.
Why Legal Work Is Hard for Agents
General-purpose agents can perform well in a demonstration and still fail in practice. Legal work is difficult for them for three reasons. First, the work is grounded: an answer that is not anchored in this matter’s documents creates risk for the firm instead of helping it. Second, the work is accountable: someone with a practising certificate signs it, so there is no room for work that is only mostly right. Third, the work is shared: clients, agents and other parties see the output, so confident errors cost the firm the trust of the people it works with.
We built Donna Agents around those three constraints.
Grounded in the Space
Every agent run happens inside a Space, the one place where a matter lives. That gives the agent access to the actual state of the work: the contract and its versions, the correspondence, the timeline, the tasks, who the parties are and what has already been agreed. Agents retrieve from the matter itself, and only from the parts of it the person asking is allowed to see. Permissions are enforced beneath the model, so an agent cannot read anything the person it works for cannot.
Verified Against the Source
The verify step in the loop is a real check. Claims an agent makes about your documents must resolve to citations, and a citation does more than name a file. Each citation opens the exact highlighted region of the exact page it came from, so a lawyer can check the agent’s work with one click. When output fails verification, the agent does not soften the wording. It sends the work back around the loop.
Approval Before Anything Changes
Donna acts for the professional, and the professional decides. Anything that changes the state of a matter in a way others can see stops at an approval gate first. That includes sending a document to the other side, marking a task as done on the client’s timeline and committing a redline. The agent presents what it did, why, and the evidence. You approve, adjust or reject it, and every run leaves a full audit trail of each step, each tool call and each source consulted.
What Agents Do in Donna Today
- Work of any size. Ask for an outcome and Donna works out the steps. A large job, such as reading a data room of thousands of pages, is shared across a team of agents working at once, then brought back together for your review.
- Research you can check. Agents work from the matter’s documents and the sources they search, and show every source they relied on.
- Work back in Word and Excel. Drafts, reviews and schedules come back as documents your firm already uses, with every assertion cited to its source region.
- Workflow Agents. Build a repeatable process once and run it on every matter, such as a disclosure review, a settlement checklist or a client onboarding sequence. The workflow records how your firm does the work, and the agent carries it out.
- Memory. Donna remembers how you work and what each client accepts, and brings it to the next piece of work.
What Comes Next
We are extending agents to carry work that runs over several days: watching for the other side’s response, picking the thread back up and keeping the timeline current in between. As the loop gets longer, the approval gates stay where they are, and the professional still reviews the work before it changes the matter.
Donna Agents are available on every plan. To try one, open a Space and ask for an outcome instead of an answer.