AI in 60 Seconds

A practical guide for lawyers who don’t want Word to “do their job,” but are happy to let it stop wasting their time. Every day, lawyers used Microsoft Word like a hammer. One day, Word started pointing things out. Because of that, the work changed. Because of that, the smart lawyers did too. Until finally… the document stopped being dead weight. That’s the shift we’re actually talking about. Not “AI replacing lawyers.” Not even close. This is about how to use Microsoft’s new legal agent inside Word without either over‑trusting it or ignoring it out of pride. Think of this less like hiring a junior associate and more like turning on the lights in a room you’ve been working in half‑dark for years. First rule: Don’t ask it to think for you. Ask it to show you things. The fastest way to misuse this tool is to treat it like an oracle. The right move is smaller and more boring. Use the legal agent to: Surface obligations you already planned to look for Flag clauses that might deserve att...

I spent months hunting for “ better prompts .” The truth? Prompts weren’t the problem—my process was. The turning point came on a week packed with meetings, follow-ups, and half-finished drafts. I kept asking Copilot for one-off help—an email here, a summary there—and I kept getting “almost right” outputs that still needed heavy cleanup. Then I tried something different: I stopped prompting for answers and started designing an AI system . Same inputs. Same steps. Same checkpoints. Copilot didn’t just help me write—it helped me run the work . The core insight: prompts are tactics. Systems are leverage. When you build a repeatable workflow, you get consistent quality without rethinking the process every time. Create a “front door” prompt. One prompt that always starts the job: goal, audience, constraints, and what “done” looks like. Separate modes. Ask Copilot to plan first (outline + risks), then produce (draft), then polish (tighten + tone). Standardize input...

I used to treat my notes like a graveyard. Ideas went in. Nothing useful ever came out. The shift happened when I noticed I was writing the same kinds of notes every day—meeting takeaways, reminders, loose thoughts about next steps. Instead of letting them sit there, I asked Copilot a simple question: “What actions could come out of this?” That one habit changed everything. My notes stopped being passive records and started becoming triggers . Copilot turned raw text into summaries, follow‑ups, task lists, and drafts—without me re‑thinking the work. The insight: Notes already contain intent . AI just needs permission to act on it. When you treat notes as inputs, Copilot can do the translation for you. Write notes for capture, not perfection. Messy is fine. Ask for actions. “What should happen next based on this?” Convert immediately. Don’t wait days to process notes. Standardize the flow. Same note type, same Copilot prompt. Close the loop. End every note w...

For a long time, AI felt impressive—but unreliable. Useful in bursts, frustrating in practice. Then I made one small workflow change, and everything finally clicked. The moment came after yet another “almost right” draft. Instead of asking Copilot to produce an answer, I asked it to decide how it should approach the work before doing anything else. That pause changed the quality of everything that followed. I stopped treating AI like a fast typist and started treating it like a thinking partner. One step of planning before execution eliminated most of the guesswork—and most of the rework. The insight: AI performs best when thinking and doing are separated. When you force clarity first, output quality jumps immediately. Think before you write. Ask for an outline, approach, or plan first. Lock the direction. Agree on the structure before generating content. Expose assumptions. Let AI surface risks or missing context early. Switch modes intentionally. Planning mode, ...

I used to jump straight to asking Copilot for output. Write this. Fix that. Summarize now. And I kept paying for it in rework. The change came during a rushed project where I was rewriting the same thing for the third time. Instead of asking Copilot to produce anything, I paused and asked it to think first —to plan, outline, and surface risks before writing a word. The result surprised me. The first draft wasn’t just faster—it was closer to “done.” Less fixing. Less second‑guessing. Less backtracking. The insight: Most rework happens because thinking and execution are mixed together. When AI jumps straight to output, it guesses. When it thinks first, it aligns. Start with intent, not output. I ask Copilot to clarify the goal and audience first. Request a plan before a draft. Outlines catch problems early. Surface assumptions. I ask what might be missing or risky. Approve the thinking. Once the approach is right, execution is easy. Separate modes. Planning firs...

I used to expect Copilot to be perfect. When it wasn’t, I blamed the tool—or rewrote everything myself. The shift happened when I stopped treating Copilot like a magic button and started treating it like a junior teammate . Smart. Fast. Helpful. But still in need of direction. Once I made that mental switch, everything got easier. I explained context. I reviewed its work. I gave feedback. Just like I would with a new hire. And the quality jumped almost immediately. The insight: Copilot doesn’t need better prompts as much as it needs better management. When you lead it well, it delivers far more value. Set the context. I explain the goal, audience, and constraints upfront. Ask for a first draft . Not perfection—something to react to. Give feedback, not frustration. “Tighter,” “more direct,” “less jargon.” Iterate together. Two or three passes beat one long prompt. Own the final call. Copilot assists; I decide. I spend less time reworking output and more time sha...