Ai legal assistant
Contract reviews are slow, expensive, and deeply manual, even in fast-moving legal teams. I saw an opportunity to shift this process into an AI-assisted workflow that helped legal teams stay compliant and retain full control over decisions.
Overview
We set out to build an AI-powered assistant that felt less like automation, and more like a legal teammate.
Legal teams are burdened by repetitive, error-prone redlining
Outdated spreadsheets and methods hinder timely decision-making
Inefficiencies delay deal closures and expose organizations to risk
There’s a persistent gap between AI’s potential and actual user trust
My role
I defined the assistant’s interaction model, designed safety nets for edge cases, and made key decisions to balance AI speed with human judgment.
business impact
60%
Reduction in redlining time of high-volume SMB contracts
25%
Of all HyperStart customers start their redlining process with Playbook
taking flight
Contemporary contracts demand swift, precise review paving the way for intelligent automation that empowers legal teams to work faster with greater confidence.
I spoke to 15 teams (120 lawyers) to identify the key tenets for our solution.
AI + human oversight—Automate repetitive tasks while users retain full control
Layered information—Show only what’s needed, with deeper context on demand
Transparent feedback loops—Let users validate AI suggestions and correct errors to refine accuracy
Seamless adoption—Integrate smoothly into existing workflows with minimal disruption
initial ideas
I mapped the journey, made detailed wireframes, covering all use cases and scenarios. Then I moved on to chart every interaction with detail, high-fidelity wireframes, ensuring no aspect of the user journey was overlooked.
value before the build
High-fidelity prototypes gave our business teams the confidence to pitch early, securing buy-in from multiple customers before we even built the product.
the core experience
And there it was the AI playbook—Instantly spot non-compliant clauses, no more sifting through endless text, rewrite or fix problem areas with a single click, without leaving your workflow, always retain the final say, override AI suggestions or confirm them at will.
Focused and actionable
AI flags non-compliant clauses and suggests fixes, information is layered to avoid overwhelming users. As the risk increases, the user goes deeper into each section with contextual options.
embedded decisions
Fallback clauses appear only when needed, with justifications, counterparty notes, and approval flows enabling structured, fast decisions. Proper hierarchy presents key details first, with deeper context available on demand.
ai summary
AI summary compiles the entire playbook run into one view, highlighting overall risk score and other details. A quick compliance snapshot, reducing analysis load on users.
quick feedback for ai
A lightweight, optional 5-point rating system at each clause review passively gathers accuracy feedback. Ensures continuous learning without interrupting user workflows, improving AI over time.
future scope
Post launch notes
After launching the pilot, we’re now working towards a self-serve model by allowing playbook configuration in workflow builder.
Monetization
I proposed a credit-based model to align usage with value, turning the Playbook into a scalable revenue layer that grows with adoption. Hopefully, the leadership will come out with v2 roll out plans soon.
Designing for AI isn’t just about connecting APIs—it’s about trust, context, and control. Especially when the stakes are high, human judgment still leads. The role of design is to bridge that gap with clarity.
There's more to this designer than meets the viewport.
This wasn't even in the wireframe. You still found it? Awesome.
Drop a howdy.
© Craftedbysom 2025
I've crafted every pixel with—
Let's get you on a desktop to see my portfolio
Wait, are you on a small screen?
// SOM
/// product Designer based in india, focused on product and interaction design for ai and saas