GitHub legal team built AI tools without coding

GitHub's legal department faced a common organizational problem: repetitive, labor-intensive work that consumed time without creating strategic value. Each new partnership deal required starting from scratch; DMCA notices demanded consistent, rapid analysis; contract review recycled the same judgment calls across dozens of agreements. Traditional solutions—hiring more lawyers or buying off-the-shelf software—are expensive and slow. GitHub Copilot CLI offered an alternative: a conversational interface that lets non-engineers describe what they want in plain language, plugged directly into their workflow, with immediate results.

The two accounts reveal why this mattered. Kasuku's contract tool worked because he could encode his own judgment—a preference for plain-language legal drafting—alongside structured access to prior agreements and style guidance, all in one repository. Geraci's legal analysis app succeeded because he could define methodology and output format in readable Markdown, letting the team control behavior without becoming software developers themselves. Neither tool required "perfect" design upfront; both grew iteratively as the team asked Copilot to do more. Crucially, both authors stress that the tools augment human judgment rather than replace it: they speed up routine analysis and enforce consistency, but lawyers remain in control of final decisions and risk assessment.