![](data:image/svg+xml,<svg display="block" id="83406992" role="presentation" viewBox="0 0 80 80" xmlns="http://www.w3.org/2000/svg"><g fill="transparent" height="80px" id="BEeNt4Zug" width="80px"><path d="M 0 80 L 0 0 L 80 0 L 80 80 Z" fill="transparent" height="80px" id="KPZzvmGG6" width="80px"/><path d="M 40.789 52.734 C 43.08 50.443 46.459 41.792 45.327 36.131 C 44.425 31.62 42.371 31.008 45.659 27.719 C 47.204 26.174 56.683 16.695 52.854 1.377 C 52.728 0.872 47.469 -2.93 48.758 4.808 C 51.886 23.573 29.534 42.364 9.798 22.628 C 2.649 15.479 5.849 1.855 4.928 0.934 C -0.033 -4.027 -4.384 15.197 8.359 27.941 C 16.512 36.094 30.066 37.01 37.801 33.143 C 42.302 30.892 46.201 42.563 37.247 51.516 C 31.539 57.224 21.868 54.453 21.752 55.501 C 21.235 60.153 32.877 60.646 40.789 52.734 Z" fill="rgb(246, 238, 247)" height="58.86800609536114px" id="N8YV9UcAu" transform="translate(13.121 10.566)" width="53.757757591804435px"/></g></svg>)
Legal Tech
Design System
Workflow Design
Redesigned core workflow experiences to reduce friction and improve task completion for legal teams—streamlining document collection, AI‑assisted drafting, and recovery flows within an existing legal‑tech platform.
Problem
The platform already supported end‑to‑end workflows for document collection, drafting, and recovery, but key steps lacked clarity: users were often unsure what to do next, where to find needed inputs, or how to recover from errors. Some flows contained redundant steps or unclear branching, which added friction, increased completion time, and created moments where users felt “stuck” in the interface. The design system was in place but still evolving, leading to small inconsistencies in patterns, states, and messaging that made experiences feel less cohesive than they could be.
Solution
I iterated on the existing workflows rather than replacing them, focusing on reducing at least one step in key tasks, clarifying each stage, and tightening how AI support appears at the right moment in the flow. I introduced clearer step structures, more explicit “next actions,” and fallback states for edge cases and failures so users always had a way forward and never hit a dead‑end loop.glean+2 At the same time, I evolved the design system—refining components, error and empty states, and interaction patterns—so improvements in one flow could be reused across the ecosystem instead of becoming one‑off fixes.
Research
I analyzed how legal teams use AI in workflows—intake, drafting, review, and reporting—to understand where clarity breaks down and which steps are most sensitive to friction. I looked at best practices in legal workflow automation and no‑code tools, paying special attention to how they present steps, show progress, and handle errors or incomplete inputs. Insights from this research informed the decision to prioritize step reduction, clearer task language, and robust fallback states over adding new features, so improvements would be felt immediately in everyday use.
Competitor
I reviewed legal AI and workflow tools that offer contract review, document automation, and matter management to understand baseline expectations for clarity, progress visibility, and recovery from mistakes. Many competing tools highlight powerful automation but still leave users guessing about what happens next when something goes wrong or data is missing. This project positioned the platform around “never lost in the flow”—prioritizing clear guidance and safe fallback states as key differentiators, not just the presence of AI features.
Planning
I started by mapping the existing flows step by step, identifying where users hesitated, duplicated actions, or needed to leave the interface to complete a task. Working as the day‑to‑day design owner, I partnered with developers and product stakeholders in daily standups to validate which steps could be merged or removed, where fallbacks were missing, and which changes were feasible in the short term. I then translated the refined flows into updated screens, states, and components—contributing new patterns (like clearer step headers, error handling, and recovery paths) back into the design system so they could be reused across the product.
Conclusion
This project sharpened existing AI‑assisted workflows rather than reinventing them, reducing steps in key tasks, improving clarity at each stage, and ensuring users always have a way forward through well‑designed fallback states. By evolving the design system alongside the workflows, the impact extends beyond one feature area: patterns for clarity, step reduction, and resilience are now available across the ecosystem for future work. The expected result is faster task completion within the platform, fewer moments where users feel stuck, and a more coherent, trustworthy experience for legal teams relying on AI to support—rather than complicate—their daily practice.
