Automating Footwork
How we're leveraging AI within our VC firm, and an invitation to come help us.
Welcome to issue #67 of next big thing.
Last month I wrote a post inspired by one of our portfolio companies’ goals for the year — “Automate the entire company.”
At Footwork, we’re of course thinking about this too.
Our approach has been to deconstruct the venture capital flowchart — find, decide, win, help, exit — and to think about how AI can improve inputs and outputs in each, such that it ultimately drives superior returns for our funds.
In addition to these five areas, there’s the opportunity to leverage AI in our internal operations to automate previously manual processes. We call this bucket firm building — for example, our reporting and communication to limited partners, internal finance function, managing our office, and other activities that may not perfectly reside in the core flowchart.
We have a very small team of 5 people — 2 co-founders that are the entire investment team, and 3 operations folks who give us leverage and focus on firm building. We’ve each been going deep with the models to run our own experiments of where AI can give us leverage, but our goal is to centralize these efforts, and to push towards the frontiers of the models’ capabilities.
We see a few paths to explore more deeply — one is what our friends at USV wrote about publicly last week in their Meet the Agents post. In just a couple of months, they rebuilt their systems to enable the firm to have AI co-workers in several areas. The system is centralized by a web application server, and connects into applications such as GDrive/Cal/Mail, Granola, and Notion to gain context. Another is to centralize with Claude or ChatGPT, connecting all our applications to it, and to spin up agents from there.
We’ve been experimenting with both approaches, but we feel like there is so much more to explore — for example, in setting up the context graph correctly across our team and for each of us individually, with the right security underpinning it, and in doing more novel work than just automating existing manual processes. We are open to other approaches as well — with Claude Code and Codex’s rapid improvements, and new agentic platforms launching daily, we suspect we will be building and rebuilding these systems not infrequently.
One thing on our mind as we automate Footwork is how to make sure we preserve what makes us human and what makes us unique. This week, one of our founders who has been meeting new investors remarked to us:
95%+ of the questions I have been asked are exactly the same in meetings. And I thought that was strange, and then [investor, name redacted] admitted on our call just then he was just asking what Claude told him to ask.
While we believe that AI can improve both our processes and our returns, we also believe that human judgment is what actually drives investing out-performance, perhaps even more so in the years ahead than in the years past.
If you’re reading this and intrigued, consider this a job description. We’d love to find someone who can help us in these efforts. You’ll work closely with me, based in our San Francisco office. You’ll learn about venture capital, have a tremendous amount of agency, a meaningful token budget to experiment aggressively, post publicly about your work, and help build an AI-native early-stage focused venture firm of the future with us. Tell us more about yourself here.
I started next big thing to share unfiltered thoughts. I’d love your feedback, questions, and comments!
👇🏽 please hit the ♥️ button below if you enjoyed this post.
Great write up 👏
Noticing the same re: in-meeting questions.
Good questions like all model outputs rely on strong context, and call transcripts are context wastelands if they sit siloed from other richer commercial context.
This "we also believe that human judgment is what actually drives investing out-performance, perhaps even more so in the years ahead than in the years past.". Is so key especially in venture funding, a lot of the blocking and tackling can be done with agents eg market sizing, but like most industries human input is what drives decisions.