Jemari Sapp
Back to Projects

Case Study

AI-Assisted Contract Generation

An AI-assisted legal workflow that transforms structured client intake information into jurisdiction-appropriate draft contracts for attorney review.

Outcome

Created a repeatable drafting process that reduces repetitive first-draft work while preserving attorney control over legal judgment and final content.

Project Type
Legal AI and workflow automation
My Role
AI workflow design, prompt engineering, system integration, testing, and product thinking
Users
Attorneys and legal staff
Status
Active development / internal implementation
Confidentiality
Some details have been generalized

System at a Glance

  1. 1Client Intake
  2. 2Structured Data
  3. 3Validation and Normalization
  4. 4Prompt and Template Assembly
  5. 5LLM Draft Generation
  6. 6Attorney Review
  7. 7Revision and Approval
  8. 8Final Contract

Overview

AI-Assisted Contract Generation is a legal AI workflow for converting structured client intake information into jurisdiction-appropriate draft contracts for attorney review.

The system is designed as a drafting assistant, not a replacement for legal professionals. Attorneys remain responsible for review, validation, legal judgment, approval, and final edits.

Business Problem

Attorneys often begin contract drafting using information collected during client intake. Re-entering information, organizing clauses, and producing a consistent first draft can require significant manual effort.

The opportunity was to create a structured AI-assisted workflow that could generate a reliable starting draft while preserving attorney oversight.

Goals

  • Reduce repetitive manual drafting
  • Improve consistency across contract drafts
  • Convert intake data into structured legal documents
  • Keep attorneys involved at critical decision points
  • Create a workflow that is reviewable, traceable, and appropriate for legal work

Solution

Structured intake information is processed, normalized, and passed into a controlled AI generation workflow.

The system creates a first draft based on client information, matter type, jurisdiction, required clauses, approved language or templates, and a defined output structure.

The generated document is then sent to an attorney for review, revision, and approval.

Human-in-the-Loop Design

The workflow intentionally keeps legal professionals involved. AI generates a draft, not a final legal document.

  • Attorneys validate legal accuracy and jurisdictional fit
  • High-impact decisions remain human-controlled
  • Outputs can be revised before approval
  • The workflow prioritizes speed without sacrificing oversight

My Contributions

  • Translated the legal drafting process into a structured AI workflow
  • Designed prompts and output structures
  • Connected intake data to document generation
  • Developed validation and review steps
  • Considered incomplete inputs, failure cases, and inconsistent outputs
  • Helped make the system understandable and usable for non-technical stakeholders

Challenges and Design Decisions

  • Balanced speed with accuracy
  • Prevented incomplete intake data from producing unreliable drafts
  • Maintained output consistency
  • Kept the system flexible across contract types
  • Designed appropriate attorney review checkpoints
  • Avoided over-automation in a high-responsibility environment

Technologies and Capabilities

Generative AI
Large language models
Prompt engineering
Structured outputs
Workflow automation
API integration
Human-in-the-loop review
Legal document generation
Data validation

Outcome

  • Reduced the work required to create an initial draft
  • Created a repeatable and structured drafting process
  • Improved consistency in how intake information is incorporated
  • Preserved attorney control over final legal content
  • Demonstrated how generative AI can support practical legal operations