Enrico SchaeferFouning Attorney, Litigator and Ousdie Counsel - IP, Tech, TRAVERSE LEGAL, PLC · Founding Partner @ TRAVERSE LEGAL, PLC | AI, Blockchain, Social Media, Internet

Lawyers: The 2 AI Skills You Must Master — Context Engineering & RAG Explained #ai

9 Oct 2025 · 2 min

Watch on YouTube

1:33

In this concise briefing, the speaker addresses lawyers on how to effectively use AI in legal work by mastering two essential practices: context engineering and Retrieval‑Augmented Generation (RAG). Context engineering involves deliberately choosing which facts, style rules, and document excerpts are placed into the AI’s limited token window, treating the prompt as a single legal‑pad page that the model can only see. RAG adds a retrieval step that searches the firm’s own knowledge base—prior briefs, client memos, local rules, discovery files, and checklists—extracts relevant chunks, and then generates drafts with proper citations, reducing hallucinations and anchoring output in proprietary data. The speaker emphasizes that without intentional prompt design and firm‑specific retrieval, AI tools cannot be reliably judged or deployed. Mastering both skills enables lawyers to command AI to draft motions using winning briefs, summarize depositions against internal checklists, and generally produce work that is accurate, defensible, and grounded in the firm’s own resources.

Chapters

Lawyers, quick PSA. If you're using AI, 2 ideas will make or break your results: context engineering and RAG. Context engineering is just deciding what the AI can see right now. Every chat has a limited token chain. Think of it as one page on a legal pad. If it's not on the page, the model can't use it. So, you choose what goes in. Your role— you are my litigation assistant. The task, The facts that matter, style rules, and any key excerpts or ideas. Uploading a document into the thread helps, but the win is being intentional about those instructions and the documents and the snippets into the thread. RAG, Retrieval Augmented Generation, does something extra. Before drafting, it searches your firm's own materials, prior briefs, client memos, local rules, your extranet, discovery. It pulls the relevant chunks, then writes with citations and references. That means fewer hallucinations and answers tied to your knowledge, not the public internet. Why should you care? Without these, you can't judge tools or build agents. With these ideas, you can say, draft a motion using our last winning brief in this court plus these facts, or summarize this deposition against our checklist, and it will go find that checklist. The takeaway here is that context engineering decides what the model sees, RAG decides where it looks. Master both and AI actually works on your data.