Why I’ve stopped trying to stay organized
Field Note #2
Welcome Back and Happy Friday!
We’ve got a long one today. I have so much I want to share. So, buckle-in!
This marks week 4 as GC of Worksome. I am a boomerang employee who took a 14 month break to teach legal teams how to use AI at Wordsmith AI. It is astounding to see how much more I can accomplish with AI now compared to 14 months ago. It is mind blowing.
So, now, I am completely changing how I work.
Firstly, I've stopped trying to stay organized the way I used to. No elaborate tracking system. No colour-coded spreadsheet of open legal matters. Instead, I'm doing something that feels almost like the opposite: I'm just documenting everything, and letting AI do the organizing.
Here's what that actually looks like in practice.
Every legal request is one of two things: a decision or process change
I've started applying a simple framework to everything that comes into legal. Any request, any question, any issue — it's either going to result in a decision, or it's going to result in a process change. That's it. Those are the only two outputs legal ever produces.
For things I can immediately answer, I answer in Slack. For issues that require more research, information or additional input from others, I create a Google doc. What the document includes:
tl;dr for the c-suite
decision needed
facts
unknowns
options (plus anticipated reactions/ramifications)
recommendation and reasoning
The document forces clarity. It also creates a record (I am calling it a LEGO Doc because they are the building blocks of the Ask Legal repository).
It also surfaces where we have gaps in processes or documents.
So let’s say one question results in two issues - one decision and process to fix like a template that needs updating, or a process that doesn't hold up at scale. So from one request, I have two outputs: (1) decision today and (2) a fix for the underlying process later. I document both in one LEGO Doc and consider it only half resolved until I fix the underlying process issue.
Simple. But the reason it's working is what happens next.
Ask Legal Slack Channel: Building a legal brain that doesn't forget
Once my LEGO Doc is complete, I upload it into Wordsmith, which is connected to our Ask Legal Slack channel. Anyone in the company can ask a question there and Wordsmith draws on the knowledge base I've been building to answer it. I've also pulled all the past questions and answers out of the Slack channel and I'm adding those to the repository, too. I am teaching the system everything legal has ever dealt with at Worksome, retroactively, but also ensuring the answer is up-to-date.
What is really cool is that I can hide the citations. So Wordsmith will provide citations to the answers so that any user can verify the answer (it is AI after all). But some of my LEGO Docs are confidential and privileged. Wordsmith has a solution for this. Wordsmith can access my reasoning from my LEGO docs without revealing the underlying source to the person asking. Confidential context stays confidential. But the logic still lives in the system. My past decisions inform future ones automatically. That is pretty awesome.
Every novel question I deal with today makes the system smarter for tomorrow. Eventually a meaningful proportion of Ask Legal questions will be answered without me. The ones that land on my desk will be genuinely novel, complex, or high-stakes. Which is, arguably, what I should be spending my time on.
Most legal functions don't operate this way. Every question gets answered fresh, from memory, by whoever picks it up. Nothing accumulates. Collective knowledge walks out the door when people leave, which, I have learned can be extremely painful.
What I'm building is a legal function where knowledge compounds. Where I'm not the bottleneck, because the answers I've already given are available without me and don’t require my business to search for answers (which we know they don’t do).
That's not a tool choice. It's an architectural decision. And it's taking shape.
Other things I built this week
I used Claude Code and Vercel to build an interactive comparison of Worksome's US features for employees and independent contractors versus our competitors. It took minutes. It's live, it’s easy to navigate, and the sales team can use it themselves without asking anyone for help. Although not technically legal work, it was a quick way to gather all the information and surface in an easy to navigate and usable way. For obvious reasons, I can’t show it to you, but I will do a deeper dive on how to do this in The Field Guide.
This wasn't a legal project. It was a business enablement tool. And it happened because the barrier to building something genuinely useful is now low enough that a GC can do it all by herself, without an engineering team or a project ticket.
I'm still thinking about what that means for how legal functions should be resourced and structured. But the short version is: if you can describe what you need clearly, you can probably build it now. That is changing how I work.
I also continued to build out the repositories for the Ask Legal channel and modify the custom instructions as both myself and the business test it with real use cases. I’ve put it into the hands of the business even though it’s not fully formed. I am focusing on continuous improvement, not perfection.
What I'm still figuring out
I need to test out the automation side of storing signed agreements and surfacing that information to the business in a meaningful way. Personally, I also never want to answer questions about a standard signed MSA again. So we use DocuSign, GDrive, Salesforce, Slack and Wordsmith. The idea is that we push the template MSA information from Salesforce into a Slack channel, which then triggers one of two Wordsmith Blueprints, which then drops the completed MSA Template back into Slack and tags the sales person. They download the agreement and send it to the client. Right now, this involves a lot of manual work.
Similarly, at the other end, we use DocuSign to sign agreements. I think I can use Cowork to schedule a task that daily grabs all the signed Agreements and then either creates a folder (for new clients) or saves the agreement to the proper folder (for existing clients). Ideally we would also run a check against the information in the agreement and Salesforce to see if it matches, but that is the step I am less confident about.
What I want to build
My Wordsmith connectors currently include Google Drive, Notion, Granola (my meeting note-taker), and Slack. I'm working on adding more through MCP. The goal I'm working toward: connecting Google Calendar and Gmail so I can simply ask Wordsmith what I missed last week, what I committed to but haven't done, what needs my attention. Currently, Gemini can do this with Google Calendar, GMail, and GDrive but Gemini meeting notes are not as good as Granola’s and only work when you have Google Meet. Granola grabs everything including in person conversations.
What this requires is not better organization on my part (I’m letting Claude CoWork do that part for me). It requires better documentation in the moment. And I've noticed our business is starting to shift too. In meetings, there's a beat now where someone says: hold on, let me make sure this is captured. Not because anyone told them to but because they've started to understand that captured information is useful information. Undocumented information disappears.
We're moving toward a model where the human job is to generate and record information, and the AI job is to organize, synthesise, surface, and answer. That's a different division of labour than most legal teams are running right now.
Legal AI vendor demos: update
At work, I use Gemini (Google Suite company), Claude, ChatGPT and Wordsmith. Because I’ve not seen another legal AI tool for 14 months, I want to do my due diligence and see what’s on the market. Is Wordsmith the best fit for me? So I will be comparing all tools against what I can do with Wordsmith, which is led by Ross McNairn and amazing GC Lucy Tyrell.
Quick Initial Thoughts (I will turn this into a downloadable report later):
I’d heard a lot about Ivo from my fellow Americans. I had to check it out. It is primarily a contracting tool. It has a cool dashboard where you can get data about not only the product usage, but more importantly about your contracts and positioning. So I really like the data dashboard.
About 1 ½ years ago, I spent a day in Stockholm with the Legora team. So I am familiar with the product and its positioning. In the beginning they focused on law firms. Now, they have expanded to serve in-house teams. So I thought I should check it out. They were really honest with me and said they are built for Microsoft organizations. Since my tech stack is Gsuite, Slack, and Notion, I should stick with Wordsmith.
I was super honest with the Harvey rep. When Harvey first came out, I tried to sign-up. I couldn’t because I was a lowly in-house counsel and not a law firm lawyer. Then I tried to get a demo but my team was less than 50 lawyers so they would not speak to me. At a conference, I tried to negotiate a 24 hour license just to see the platform. So it was not until February 2026 that I actually was “allowed” to see Harvey. It was better than expected based on what I have heard. However, they just released a connect with Claude Cowork, so Harvey got a lot more interesting.
Cecilia and I have swapped LinkedIn messages over the last several years. Prior to Wordsmith, I used GC AI because it was the only tool at the time focusing on in-house counsel, which was awesome. The platform has evolved a lot over the past 14 months. I am interested in testing its research capabilities as I have heard good things about it.
I like to consider myself LinkedIn friends with John Lindsey, who is an amazing human being and VP of Sales at Term Scout. When he reached out about Term Scout I was a little skeptical. But after hearing more about the product, I was intrigued. Term Scout is not like AI legal tools. It has a different and complementary value proposition that is ahead of its time (probably because it’s led by Olga Mack). More on this later.
Built by Legal Ops and CLOC superstar, Jenn McCarron, I was interested to see how her company uses contracts to surface data. I was going to ping Jenn McCarron for a demo but it turns out you can just sign up for free on the website — which is exactly the kind of low-friction entry point I appreciate. More to report once I've actually used it.
The pattern I keep seeing: the highest-value automation in legal isn't the glamorous AI-drafts-a-contract use case. It's the boring connective tissue that nobody wants to do and everybody forgets (but gets me excited!).
A quick tool note: voice
For voice capture, I'm using ChatGPT rather than Claude. The voice transcription is noticeably better. I use Claude for a lot of the heavy analytical lifting, but when I'm capturing a thought on the move or at my desk, ChatGPT is always better (which makes me sad). Use what works. There's no prize for tool loyalty. If you want to follow the captain of voice in legal tech, reach out to Andrew and tell him I sent you.
*Proof-reading has NEVER been a strong skill of mine. So blame Claude if there are typos.
If you have questions or want to follow up on anything, hit reply.
Want the Field Notes to hit your inbox each week? Subscribe here.
Want to dive deeper? Become a member of The Field Guide.
