What I Learned Being a Chief of Staff

TL;DR

Ask your principal for the company’s top priorities and act on them instead of waiting for work to be delegated.
Earn a say in strategy: when taking on something new, you'd be surprised how much you can learn within a couple of weeks of reading and talking to people who’ve done it.
Look for leverage: a few days of your prep, like hosting events and briefing your principal on who'll be in the room, can create a lot of value for them.
Over-communicate by sending your principal a daily list of what you plan to work on, rather than staying silent and leaving them to guess.
Identify the critical path and follow it, lest you waste effort.
Having a weekly reflection that includes growth areas and a request for feedback will accelerate your growth more than passive on-the-job learning.

Intro

This summer, as part of the Generator Residency, I spent nine weeks as chief of staff (CoS) to someone launching an early-stage startup that automates analysis of the strategic implications of new technologies.

I initially thought the role would be doing admin and creating systems to save my boss work, and that owning a workstream or being a strategic partner would come much later. This guess was incorrect. I took on the hiring workstream in the first couple of weeks before moving on to aid in fundraising.

Every chief of staff role is different; it varies based on the comparative advantage between you and your principal, as well as the company's stage. In this post, I will share the most important lessons I learned while being CoS that I believe made me much more effective.

Ask for the top priorities, then act on one

I spent my first week gaining context about the company and taking on smaller tasks. That work was useful, but it wasn't directly addressing the company's biggest problems, and that's where I wanted to be. After a week, I asked my principal about the organization's priorities. He listed fundraising first and hiring second. He was already doing the fundraising, so I took hiring even though I'd never hired anyone.

I set out to own the hiring function. I spoke with hiring managers from METR, Kairos, and Redwood Research. I reached out to an AI safety investor and founders I'd met at events to discuss their processes. I spoke with Claude, read people’s takes online¹, and read Who: The A Method for Hiring.

In parallel, I was building out our hiring process. For both roles we were hiring for, I created a scorecard built backward from what the role needed to produce, a detailed job description that filtered out candidates who obviously wouldn’t be a good fit, and a candidate tracking system in Airtable. I scoured candidate databases, reached out to those I thought were promising, and set up calls with them and my principal.

That's the payoff of owning a priority: you can move much faster, while your principal can focus on the most important function in the business.

Earn a say in strategy by building a gears-level model

After a few weeks, our hiring process was functional, and the bottleneck had shifted: we needed funding before we could put people through it. My principal had a good handle on philanthropic funding, so my new objective was to determine whether VC funding would be a good fit for us, how we would secure it, and the most likely reasons we would get passed over.

The answers weren’t simple: there is a lot of noise in the VC funding space. Instead of listening to LinkedIn gurus for advice, I just did the reading.

I read internal memos from VC firms, surveys investigating how venture capitalists make decisions, and guides written by investors and founders². From these readings, I built a model of how VCs work and could use it to make higher-level strategic decisions:

We could afford to delay thinking in detail about the startup’s governance structure until the Series A round. This is because the SAFE does not grant a board seat.
At our stage, finding a technical cofounder or founding engineer was a priority. At the pre-seed stage, the single biggest factor in VCs' decision-making is the quality of the founding team.
We should think through and address the hardest questions in our initial pitch to a VC. After the initial meeting with a VC, the VC meets with its partners to discuss whether to proceed with due diligence. Therefore, the first meeting has to get the VC excited about you and arm them with answers to the hard questions their partners will ask.
We should focus on building the team and product, rather than working out a detailed business model. Valuations come mostly from the expected exit and comparable companies rather than a cash-flow model, and at the earliest stage many VCs don't forecast at all.³

If you want a say in strategy, go learn the part of the problem your principal hasn’t.

Create luck

During the final couple of weeks of the Generator Residency, a whole week of tech events was happening in DC: a great opportunity to find cofounders, hires, and investors. The right person could change the company's direction.

My goal, therefore, was to maximize our luck. I did this in two major ways: hosting events to attract the right people and briefing my boss on the most important people at the events he was attending.

For the former, I decided we should host two events: one to attract people likely to understand and share our thesis, and another to attract startup founders and investors. I came up with and ran the events end-to-end in 2 weeks.

For the latter, I scraped the Luma event pages to compile a list of all attendees and created briefings for the most important people, so my principal could bump into the right person and know what to say.

Fewer people came than we expected, but the right ones did. My boss also met several people from my briefings that he wouldn't have found on his own. It's too early to say what the week will lead to, but we left with a few very good contacts.

Show your boss what you're doing and ask for help

For most of the summer, I tried to stay out of my boss's way. He was the most expensive person I could go to for help, and I worried about taking up more of his time than the work required. Early on, I spent two and a half hours stuck on a draft for 80,000 Hours' candidate search because I didn't want to take his time. Asking him would have saved me about two hours.

I was concerned about being high-maintenance, but the real problem was that I didn't communicate what I was working on or where I could use help.

The solution to this was extremely simple: send him a list of what I intended to work on that day, roughly in order of priority. This way, my boss could see what I was working on and give me tips on how to approach it.

Own your own growth

I chose this role because it was the best opportunity for me to grow. I did grow a lot; however, this growth started out mostly passive.

I would grow by getting feedback from my boss on my work and gaining experience through learning and doing, but it wasn't as fast as it could have been. My growth curve began to rise when I scheduled weekly reflections with my principal. I used this 1-on-1 template from Max Dalton, which notably included asking for feedback and making my growth areas explicit. This simple system allowed me to collect much more feedback, targeted at the areas where I wanted to grow the most.

Example: Prioritization

One area I'm strengthening is prioritization while executing. I was able to take on the right work, but the order in which I completed the tasks was not effective.

Two examples:

I researched and designed work tests and structured interviews before the job application went out and before anyone went through our process. A work test is useless without applicants, and the one thing that would get us applicants was the application.
I spent too long on AI-generated images, a cooler website, and the perfect wording, rather than just posting the event online and sending out invitations.

The generalizable lesson is to identify and follow the critical path.

Final thoughts

I enjoyed being an important part of the business function, and the value I provided was obvious. I'm certain that getting the hiring pipeline moving, sending outreach, and organizing the events were counterfactual. I, however, will not be continuing in this role for the time being, because I want to explore other roles that might fit me better.

I'm early in my career, and there may not be much time before transformative AI redefines much of society, so I'm focused on finding the role that best fits me. Right now, I'm exploring moving deeper into recruiting, hiring⁴, AI enablement, and special projects roles.

Thanks

Thank you to Alexandra Bates, Director of Convening at Constellation, and Tzu Kit Chan, former Chief of Staff at Atlas Computing, for mentoring me throughout this. Alexandra improved my personal effectiveness by helping me navigate my emotional blockers. Tzu gave me specific advice on how to do the work of a chief of staff well.

Thank you to Ivan Shishkin, Tzu Kit Chan, and Aris Richardson for their feedback on drafts of this post.

What I read on hiring, besides Who: Sam Altman, How to hire. How to Hire a Team on LessWrong. Some surprising hiring practices I follow on the EA Forum. Sam Rodriques, You can hire people who know more than you do. Mechanize, How we interview software engineers. Graham Duncan, What's going on here, with this human? ↩
NFX, How VCs Decide to Take a First Meeting. Everything I learned raising my first pre-seed round. Lenny's Newsletter, Raising a seed round 101. On SAFEs and how a round works: Y Combinator's Understanding SAFEs and Priced Equity Rounds and Modern Startup Funding. On the partner meeting: First Round, Here's What You Can Really Expect When Pitching Your Seed-Stage Startup at a VC Partner Meeting. ↩
Paul Gompers, William Gornall, Steven N. Kaplan, and Ilya A. Strebulaev, How Do Venture Capitalists Make Decisions?, NBER Working Paper 22587, 2016. Published in the Journal of Financial Economics, 2020. ↩
This fall, I'm mentoring a SPAR project on hiring resources for new AI safety organizations. ↩