Introducing Agent Builder: Build Smarter Agents for Complex Legal Work

Harvey is agentic by design. Our platform is built to handle multi-step tasks, pull from different tools, and adapt findings along the way. Across Harvey, there are numerous agents that help fulfill different tasks across workstreams, such as detailed analysis, drafting, and term extraction.

Agents are critical to how lawyers use Harvey today:

• Our platform receives 400K+ agentic queries daily to complete legal tasks like contracting reviews and document editing.
• Teams use agents in Harvey to extract 20M+ terms in review tables, helping them streamline document reviews across litigation and transactional use cases.
• Users have created 445K+ reports with agentic Deep Analysis.

Additionally, legal teams have already built over 25,000 agentic custom workflows with Workflow Builder to structure and automate recurring tasks. With this model, teams could define their inputs, set their steps, and let Harvey produce consistent, reliable outputs. Teams like GSK Stockmann have built their own agentic workflows to speed up due diligence reviews. Ashurst has also built agentic workflows for specific practice groups; in their experience, their team saves hours on producing lease summaries.

As the model frontier continues to evolve, improvements in the underlying models open up new opportunities to scale their expertise. That's why we're evolving Workflow Builder into Agent Builder.

Watch Video

Build Agents That Reflect Your Organization's Processes

Agent Builder helps users easily create their own Workflow agents tailored to specific teams, practice areas, or internal processes and then share them with others. Teams can embed templates to provide additional context on how the Workflow agent should produce work, ensuring that each output is useful. Once built, those Workflow agents are reusable, which can help teams get through new tasks faster.

Lawyers can already see suggestions for specific Workflow agents when they’re working in Harvey’s Assistant today. In the future, users will also see relevant custom Workflow agents when they align with an ad-hoc prompt, making it easier for them to utilize tools their teams have already set up.

Looking ahead, teams will also be able to schedule Workflow agents to run in the background. This ensures that Workflow agents can finish tasks more autonomously, and that they can parallelize execution without the user needing to step in. Scheduling can be useful for tasks like monitoring contract expiration dates or running regular compliance checks, so that the Workflow agent can summarize initial findings without waiting for someone to manually identify what actions need to take place each time.

Handle Multi-Step Tasks More Efficiently

Workflow agents take fuller advantage of the latest model reasoning improvements. They can now accomplish more with less detailed instructions, infer what's needed, and navigate tasks more efficiently. Rather than rigidly moving through a fixed set of steps, a Workflow agent grasps the goal, asks for the right context, and finds the most direct path to a useful outcome while still ensuring humans are in the loop.

In practice, this might look like a transactional team using Workflow agents as they run through a due diligence review. Other teams can use a Workflow agent to search over a wide set of documents, build detailed review tables that surface core terms and material risks, and then immediately create a diligence risk summary or post-closing checklist. The Workflow agent can iterate based on feedback to ensure that results align with expectations.

For a litigation team, it could mean starting out with a detailed review of a set of documents, and then using that review to automatically generate a high-quality first draft memo. The Workflow agent connects those steps in a single, coherent process without needing to break up work into multiple steps.

For legal teams, this makes a lot of the manual work that goes into routine tasks more effortless. Instead of writing exhaustive instructions to cover every possible scenario, lawyers can shift their time towards setting the objective and letting the agent work toward it — while staying involved where it counts.

Stay in Control as Agents Complete Work

More autonomous work requires more deliberate oversight. Workflow agents are designed with human-in-the-loop checkpoints as a core feature, so that lawyers stay in control of the decisions that matter. As an agent moves through a task, it surfaces decisions and flags moments where pre-defined critical user inputs would improve the results.

For example, an in-house legal team that’s responsible for reviewing regulatory updates may use a Workflow agent to search through a set of existing company policies and flag language that no longer aligns with recent regulatory updates. The Workflow agent can then ask the user to confirm whether the right approach was made and whether the Workflow agent should proceed with editing the documents directly. This ensures work keeps moving and that lawyers can guide the work in real time, with the visibility to course-correct where needed to align with their judgment.

Focus on Driving Better Outcomes

Agent Builder helps give legal teams a foundation for scaling consistent, high-quality work — one that adapts to their processes, rather than the other way around. This can help legal teams stay focused on the tasks that lead to better results and client relationships.

Looking ahead, we’re continuously investing in improving agent functionality across the Harvey platform. This includes incorporating long-horizon agents in Harvey to help teams execute complex tasks like fund formation, as well as developing tailored agents in Shared Spaces to improve collaboration across teams.

If you're already using Workflow Builder, your existing workflows remain fully supported with no changes to existing functionality.

If you’re interested in learning more about how Agent Builder works in Harvey, reach out to your account team or contact us: