GitHub legal team builds AI tools without coding using Copilot CLI

GitHub's legal and business team built a series of internal AI tools using Copilot CLI, a feature that accepts plain-language requests to generate and customize workflows. The team—lawyers, program managers, and business professionals—faced common pain points: repetitive tasks like reviewing similar contracts and answering recurring legal questions. Using Copilot CLI, they discovered they could build solutions without hiring engineers or waiting for external vendors.

Ngandu Kasuku, a product attorney, initially used Copilot CLI for one-off tasks but realized he could build something larger around how he actually worked. He created terms-ai, a contract drafting tool that organized key documents, instructions, and workflows in a single repository. The tool includes an internal drafting style guide based on plain-language legal principles—Kasuku had long favored simple language over archaic terms like "heretofore" and "therewith," and discovered an entire legal drafting movement shared this view. The tool also accesses a library of previously completed agreements, allowing it to draw on approved, access-controlled templates when partners send new agreements or addenda. Since building terms-ai, Kasuku has cut his review and drafting time roughly in half, and his provisions are now more consistent across agreements. The tool and its general workflow are open source, though sensitive agreements are kept in an internal repository.

Jesse Geraci, an online safety counsel, started with a different problem: analyzing source code quickly and accurately to evaluate DMCA (Digital Millennium Copyright Act) notices. His original project began as a set of GitHub Copilot instructions for recurring tasks—DMCA triage, code comparison, license checks, and circumvention review—that aimed to turn one-off prompts into something a legal team could trust. Rather than writing source code, Geraci used his legal skills to build plain-language instruction sets, policy references, and report templates. The workflow grew to include different analysis modes for clients (faster outputs with escalation recommendations) and lawyers (deeper review with both-sides arguments), as well as integrated external data sources. When he handed it to the team, they immediately adopted it and asked Copilot to do more. The foundation evolved into a full desktop app with a clean interface for running predefined legal workflows. Although the app required writing some code, the core instructions—which customize behavior and control workflow routing—remain written in plain, editable Markdown. The app now handles many in-house workflows beyond DMCA code analysis: contract review, NDA triage, risk assessment, compliance checks, and response drafting. Under the hood, it routes work through reusable skills and agents (intake, playbook alignment, risk scoring, evidence verification, escalation routing, report assembly), but legal teams control behavior through readable Markdown rather than engineering changes. Geraci emphasized that his tool is a structured decision-support system designed to keep human review central while making legal analysis more consistent, transparent, and scalable—not a replacement for legal judgment.