I am, by any reasonable measure, an organized person. My calendar is color-coded. My Notion workspace has a Notion workspace. I have a shelf of small Muji notebooks filled with my hand-written notes, because when I write something down by hand, it stays in my head. There is something about the physical act of committing ink to paper that moves information from the short-term pile into something more permanent, and I have trusted that mechanism for my entire professional (and academic) career. I use Granola now to capture meeting notes, which is a concession to speed I have made and do not regret. The notebooks remain, a silent analog anchor that I am not quite ready to cut loose.

What I am not, and have never been, is a coder. What I am building anyway, with Claude Cowork and no small amount of stubbornness, is a Chief of Staff.

Most people who hire a Chief of Staff are trying to reclaim their calendar, but I already have control over my calendar and to-do lists. What I needed was something closer to institutional memory: a system that carries context forward, travels with me from conversation to conversation, and never makes me reconstruct a client relationship from scratch at 8am.

There was a call last fall where a client mentioned offhandedly that her team had just lost their biggest internal champion for the project (someone she had referenced by name in our very first conversation), and I had nothing. No note, no time to shuffle through my notebooks, and no memory of the name at all. I moved through it, but I felt the slip. On a good day, I had the notes. On a real day, I had the notes, plus three other calls, a proposal in progress, and a class I was teaching, and the connective tissue between client conversations was quietly fraying. My tools were not talking to each other, and I was the one always doing the translation.

I needed something that would show up thirty minutes before a call and say: here is who this person is, here is where you left things, here is what to listen for today. Something that would sit with me afterward and make sure nothing important slipped through. I knew exactly what I was describing because I had done this work myself. I have been a Chief of Staff, and I was a good one. The function I was missing was the one I used to provide for someone else, and It took me longer than I would like to admit to realize that I needed to build it for myself.

The system lives in a folder named "Chief of Staff". It has four commands. Before any client call, one command produces a one-page brief: relationship history, open items, what to listen for, assembled from my call transcripts, my client file, and whatever is publicly relevant to that person right now. After the call ends, a second command pulls the fresh transcript, extracts decisions and commitments, and the quieter signals about how the relationship is actually going, updates the client record, and drafts a follow-up email that I review before anything goes anywhere. By design, the system never sends. Two other commands manage my mornings: overnight inbound triaged by urgency, routine responses drafted for my review, judgment calls flagged for my attention. A fourth scans what my audience is actively discussing so I can surface the topics people are asking about, so I can write about them or do my own tests.

The system knows my clients. It knows my voice. It knows which work requires my judgment and which is pattern-based and can be prepared in advance.

Anthropic publishes a library of pre-built skills and plug-ins: structured knowledge files that give Claude specific capabilities right out of the box. I didn’t have to build my Chief of Staff from scratch. I took the ones that applied to my practice, read through them carefully, the way I would read through a brief before reworking it, and then customized them to reflect how I actually operate: my clients, my communication style, my standards for what a good debrief looks like versus a lazy one. The skills provided the architecture, and I provided the context that made the architecture mine.

The conversation about who can build AI systems is changing. The assumption was that it requires a developer's fluency, and for certain things, it still does. For my system, what it required was the ability to describe a process in precise language and the domain knowledge to know what a good output looks like versus a mediocre one. You are writing creative briefs, essentially, and every skilled practitioner already has that raw material. Most just have not realized that the gap between having a workflow and having a working tool has gotten dramatically shorter.

Every output the system produces is a preparation document for my review. The follow-up email drafts sit in a queue until I read them and decide what to send. The morning review flags things; I make the calls. The content signals surface what my audience is asking about; I decide what is worth saying and how to say it. The newsletter stays entirely mine, as does the strategy, and anything that requires taste, experience, or accountability in a client relationship. The system exists to make sure my judgment is well-informed when I need it. I spent years doing this for someone else. It turns out I needed to do it for myself.

If you want to talk through what an intelligence layer could look like for your own practice, or if you want help building one, let’s chat.

— Lauren Eve Cantor

thanks for reading!

if someone sent this to you or you haven’t done so yet, please sign up so you never miss an issue.

I’ve also started publishing more frequently on LinkedIn, and you can follow me here

if you’d like to chat further about opportunities or interest in AI, or this newsletter, please feel free to reply.

banner images created with Midjourney.