I completed the Pivotal Fellowship in Q1 and have been fielding questions from people interested in similar fellowships—particularly those early in their careers or considering a switch into AI policy. I thought I'd write up some rough reflections. I'm timeboxing this to two hours, so it's not exhaustive and might have some sloppy writing, so I'm happy to answer any questions or fix things.

So what did I actually do?

I received my fellowship offer in December, with the programme due to begin in February. During the weeks leading up to the start, I worked with my research manager (RM) to figure out what direction I wanted to explore and who might serve as a good mentor. With my legal background, I knew I wanted to work on liability and tort law for AI labs—particularly within a UK context.

This 'pre-fellowship' period involves extensive mentor matching. Whilst this is no longer the case with Pivotal (you now apply directly to a mentor), programmes like ERA still involve onboarding a mentor during this phase. You'll spend the run-up period figuring out who could best serve your research needs. Your RM typically helps sort this out, though you'll also need to provide useful context about what you're looking for.

I had about three to four people who seemed like good options but weren't available, and eventually found someone suitable near the start of the fellowship. My mentor and I discussed what kinds of questions would be exciting to tackle—he gave me several papers to read whilst I scoped out specific subquestions I wanted to address.

Weeks 1-3: Orient

The first few weeks are largely about orientation. This includes adjusting to your new environment—for me, that meant moving to London, familiarising myself with the new office, and meeting the other fellows. It's quite something, the new world that opens up to you.

Research-wise, I spent weeks 1-3 writing out subquestions and outlines. You simply cannot answer everything you want in nine weeks, so you need to get quite specific and narrow your focus. Through weekly meetings with your RM and mentor, much of the time is spent doing this narrowing and understanding what you actually want to accomplish.

I reached out to people beyond my mentor and research manager—experts in AI governance and law. I'd typically send them a short document outlining my current thinking and explaining what kind of feedback I was seeking, usually writing something like: "This should take 15–20 minutes to read—any comments would be super helpful."

I'd then combine that feedback with my own reflections and bring it to my mentor. We'd use those conversations to make sense of what felt like a messy, ambiguous space and slowly move towards something more concrete. These weekly mentor check-ins were absolutely central to my fellowship experience. 

Meanwhile, my research manager acted more as a sounding board for execution—checking that my research process, time management, and overall trajectory were holding together. They weren't a subject matter expert on my topic, but they'd often have helpful thoughts and could flag when something seemed off-track. Essentially, an RM helps you execute and provides the backbone to your work. It was also valuable having someone check in on how I was feeling about the fellowship—though there's also a Community Health Manager specifically there to ensure you're doing well during the programme.

And then there are the fellows themselves—which is half the fun! So many brilliant people to add to your life, all in similar positions, trying to figure out their research, their careers, their lives in the age of AGI. Meeting the other fellows was perhaps the best part. You're all seated near each other during the fellowship, you socialise outside the office, you go through it all together—the banter, the stress, the feedback on your work. It really is fantastic.

The first few weeks also involve lots of talks. Many people working in AI safety or governance have done similar programmes, so they tend to be quite charitable with their time—doing short talks about their work followed by Q&As. This is incredibly helpful for gaining context quickly. It's surprising how much information you can absorb in such a short space of time

Weeks 3-6: Drafts

I think of this period as primarily about writing first drafts and circulating them for feedback. Your work won't be fantastic at this stage, but you're actively encouraged to seek feedback early rather than waiting until you have a polished piece. This inevitably means doing loads of redrafting—the point is that you should be nimble and responsive to input.

You also get involved in something called Projects in Progress (PiP). The idea is that you and your fellow researchers give each other peer feedback. The format involves three fellows being chosen to discuss their work, either through presentations or by circulating drafts. Other fellows then leave comments, followed by a discussion session. This was surprisingly helpful on multiple fronts.

For me, working on quite a niche topic, these sessions helped me learn to explain my research better. Liability law is pretty convoluted for non-lawyers, so I needed to be exceptionally clear about where my writing wasn't landing or where I needed to explain concepts more thoroughly. I also got significantly better at giving feedback—learning to be constructive rather than just blunt, and developing an eye for which comments would actually prove useful to someone else.

The usual rhythm of talks and check-ins continued throughout this period, with the occasional dash of novelty. We took a trip to Oxford to visit the GovAI fellows, which was brilliant for connecting with those researchers and getting additional feedback on my work. Oxford was also home for me, so it was lovely to visit again!

Weeks 6-9: Sprint

Things start to ramp up considerably at this point. On one hand, you're trying to finish your work—I was ideally aiming to have a paper ready by the end of the fellowship—but you're also thinking about applying to opportunities that have emerged. My head was pretty much down during this period. I got quite absorbed in writing my work well, re-reading drafts, and checking with my mentor about whether what I was saying made sense.

Because I was so focused, I honestly can't remember this time super clearly. I didn't feel stressed, but the passage of time felt much quicker—perhaps because the end was in sight and I'd become a bit tunnel-visioned on reaching the finish line with a completed paper. Which I did manage to do!

My Takes on Doing One of These Fellowships In No Particular Order

1. Don't feel pressured to finish everything. One thing I wish I'd known is that you don't actually have to complete your research by the end of the fellowship! Most importantly, your research serves as a vehicle for upskilling and gaining context through extensive feedback. Most people finished their projects post-fellowship. The most valuable thing you can do is reach out to loads of people. Being able to say "I'm a Pivotal fellow" (or whatever relevant fellowship) opens up doors and lets you speak with far more experts than you'd normally access.
2. Okay, but you should also be prioritising your research—a thing to avoid doing here, however, is having a default idea of what the output would look like. It doesn't have to be a paper on ArXiv, SSRN, or a blog post. It doesn't have to be a traditional academic paper at all. Perhaps you want to send a memo to MPs, or create an instruction manual for policymakers. The key is figuring out who your audience is, what would be most useful to them, and then crafting something that actually serves their needs. Even if you are new to this whole AI thing, your research can genuinely help someone—identifying who that someone is makes all the difference.
3. Apply to things during the fellowship. I think I should have applied to more opportunities whilst I was there. As I mentioned, I was quite zoned in, and there were loads of opportunities I didn't feel qualified for. I thought once I'd finished my research, I'd be a more 'suitable' candidate with a high-quality piece of work to link in applications. This was wrong—you should still apply even early in your fellowship. Something, something, you miss all the shots you don't take.
4. Make the most of office life. Being in an office space (LISA, for me) was incredibly helpful. Aside from the endless Huel, chatting to people over lunch about what you're doing gives you far more perspectives on your situation and work. This is extremely underrated and, of course, not something you get if you're outside these environments.
5. Make the most of your fellow fellows. You have around 15 other fellows—this is an opportunity like no other to get loads of feedback on ideas beyond your research. I started a blog during the fellowship but didn't write much on it during the fellowship because I was, again, too focused on my main research. I wish I'd done more external thinking and written about my takes on DeepSeek, UK AI policy, and other topics, then circulated them to the fellows, my RM, and others at LISA for feedback.

Good luck if you're doing one of these! May the power of the fellowship be with you.