In-house attorneys spend nearly one-third of their careers directing traffic.

That’s 12 hours—one and a half days—every week.

Answering quick questions in Slack.

Chasing down missing information for contracts.

Manually assigning tasks.

Searching for past precedent.

Countless hours spent on tasks that are easily automated.

It causes burnout, frustration, and unnecessary headaches.

Imagine what you could accomplish if you regained several hours each day.

Get your time back and increase your team’s output with this comprehensive guide to improving legal team efficiency.

And if you’re curious, benchmark your team’s current efficiency with our Legal Efficiency Score calculator.

You probably already know AI can help you redline a contract. You've likely even gone further, getting it to redline more like you, using your team's playbooks, preferred language, and risk tolerances.

It's a great start. The harder question — the one most frameworks skip — is what comes after that. What else? And where do you stop?

That's what this issue is about.

This is for in-house lawyers who want to use AI and understand they need to — but aren't sure what a coherent approach actually looks like. And for those who got excited when they saw what one tool could do, and are now trying to figure out where to go from there.

I'll share a framework for thinking through effective use cases for yourself and your legal team, including where to draw lines about what AI isn't right for (at least yet). I'll also share some of what my team and I have done so far — what's worked, and what the impact has looked like.

One note before we get into it: most issues of this newsletter draw on something I've been thinking about for many years. This one doesn't. AI is newer, and I'm newer to it. I got my first paid ChatGPT account in January 2024. We've rolled out legal-specific AI tools at work more recently than that. I'm still figuring this out, too. I'm sharing the framework I've been using because it's helped — not because I have all the answers. And I know that the ways I’m using AI in six months will probably look very different than the ways I’m using it today. Just yesterday, Anthropic announced twelve new legal practice area plugins and new connectors. Here’s a summary from Legaltech Hub.

Let's get into it.

For years, the aspiration for in-house lawyers was to become a true "business partner.” This meant earning a seat at the table, getting into decisions before they were already made, and pairing legal advice with business context. That's still important. It's also no longer enough.

The next phase requires something different: being a business builder. Not just additive to someone else's agenda, but having a real stake in what the company creates. Asking "what should we build and how?" alongside "what should we avoid?" Measuring success not only by what didn't go wrong, but by what got made.

That shift matters here because of what AI is already doing. It's moving into the partner lane by surfacing risks, reviewing contracts, answering questions at 2am without ego. It's not doing it perfectly yet, but it's improving fast. The partner-lane work that used to require a lawyer's time is increasingly something AI can handle, cheaper and without the bottleneck.

What AI cannot do is build. It can't hold the judgment you've earned from years of knowing your company, your CEO, this specific moment. It can't walk into a room and shift a conversation. It can't have a real stake in what gets created.

So the question of how to deploy AI in your legal department is really a question of: what should I be spending my human judgment and limited time on? Start from your goals and work backwards.

Ask these four questions:

That last question is important to be honest with yourself about, because it can save you a lot of time and energy now. The answer will be different for every team, but the reasons tend to cluster: AI isn't yet good enough at the task relative to how much is at stake. The work involves data you've committed, whether contractually or otherwise, to keep out of certain tools. Or the human judgment and the relationship aren't separable from the output. If using AI isn’t the right approach to a task, no matter how many different prompts you try, agents you build, or AI native tools you try, the result won’t be great.

Knowing where you won't go is as useful as knowing where you will. And in most cases, what you're protecting is exactly the builder-lane work: the things that require your presence, your history, your stake in the outcome.

If you want to go deeper on this, including what can happen to your skills when AI handles more and more of the work, and how to make sure you’re still protecting what matters, I covered that in the April 15 issue.

Here's how it looks for my team — what's made us faster, what's made us better, and what's cleared mental space.

Where AI has made us faster

Contract review is the obvious one, and it's meaningful. We use GCAI with our own templates, playbooks, and preferred language loaded in — which means turnaround is faster and the output already reflects how we actually negotiate, not generic market standards.

Repurposing and templatizing used to eat time. It was more efficient than starting from scratch, but now we can move even faster. Taking a document built for one purpose and adapting it for another, like updating a board consent or repurposing a customer contract, is exactly the kind of work that's technically straightforward but burns an hour you don't have. Now it doesn't.

We've also gotten faster at intake and routing. When work comes in, AI can categorize it and get it to the right person immediately — no sorting, no lag, no work sitting in a queue waiting for someone to figure out whose problem it is. We use Streamline AI, and in addition to helping us collect clear data on volume of work, types of work, and speed, it helps us move faster. And we're taking that a step further: for certain categories of work, we're building toward AI doing the first pass entirely, so that a lawyer is reviewing and approving rather than starting from scratch. The goal isn't to remove lawyers from the loop. It's to make sure their time is spent on the part of the loop that actually requires them.

I've also built out different prompts for different email recipients. When I'm writing to a director-level person, they need enough context to act without coming back to me; they're not in the weeds day-to-day, and often care most about the corporate governance aspects of a topic. When I'm writing to certain executive colleagues, I want something different: short, minimum detail, the primary point, and often my clear recommendation. Having those prompts ready means I'm not rebuilding them from scratch every time. Scroll to the end of this newsletter for a link to versions of these prompts you can borrow for yourself.

Where AI has made us better

Some of the most useful work I've done with AI has nothing to do with drafting. Before I take a position to an executive or into a negotiation, I'll sometimes test it first, by giving AI the argument and asking it to push back. I’ll have it tell me what I'm missing and predict what the other side will say. It's a low-stakes way to stress-test reasoning before the stakes are real, and it's surfaced gaps I would have missed.

The other use case I've come to rely on is as a final check on detail-intensive work — regulatory filings, complex agreements, anything where you've been staring at the same document long enough that your brain starts auto-correcting errors it should be catching. AI is good at this. It doesn't get tired. It doesn't skim. It will find the defined term you used inconsistently on page 14.

My business-forward email prompt has also become a reflex. I give it my draft and ask it to make it more business-forward, and less centered on legal issues or risk. It doesn't change what I'm saying. It changes how I sound saying it.

Where AI has reduced the mental load

This is the one where I've seen the most impact personally as a GC, and where I think most legal teams haven't pushed hard enough yet.

By the time I pour my first cup of coffee on a weekday morning, my triage workflow has already reviewed my overnight emails and summarized them into a daily Slack canvas for me, highlighting what matters most, and what I might have missed from the day before. Rather than starting the day by manually sorting through what matters, or trying to remember what I noticed late the night before and told myself I’d prioritize in the morning, I'm starting with a map.

Monday mornings are great now too. Most in-house lawyers have made peace with starting the work week by reconstructing where everything stood — did you send that non-urgent Saturday night email, schedule a response for Monday morning, or just mean to and not? It's friction most of us accept as part of the job. I'm not accepting it anymore, and you don't have to either. Now a workflow runs Monday morning and has a summary of the legal team's progress against our weekly goals waiting for me first thing. The week starts with context instead of catch-up. Making the transition back feel less jarring is not a small thing.

I’m experimenting with other use cases too. Between Claude Cowork and Microsoft Copilot, it’s easy even for the non-technical among us to build agents (or agent-like tools) that make our lives and jobs much much better.

Individually, none of these is dramatic. Together, they've been the difference between feeling behind and feeling like we're running the department instead of the department running us. We're supporting a startup that builds startups — the volume only goes up, and the headcount doesn't. These workflows are what make that math work.

The framework isn't a checklist to complete all at once. Instead it's an approach to more fully thinking through how best to use AI for your specific legal team. Pick one question that feels important for your team right now — how to get faster, better, or lower mental load — find one use case, and build from there. This isn’t a race and, with the speed AI is moving, there’s no such thing as perfection. What we all need to do is get started experimenting and iterating.

But keep sight of what the whole thing is in service of: being a better builder. If you get faster, better, and clear the mental overhead, you’re better positioned to deliver real gains for your company; and those gains compound.

I put together templatized versions of some of my highest impact prompts and scheduled tasks.