A Five-Person Venture Firm Runs on Agents, and the Analyst Role Is the First Casualty, Footwork's Trivedi Argues

45 minutes ago

4 min read

What's New

A venture firm that captures every conversation and decision into a context layer its agents can read turns judgment into a compounding asset, and firms too large or too entrenched to rebuild will not keep up. Nikhil Basu Trivedi, Co-Founder and General Partner at Footwork, argues this in a podcast on How I Invest. Footwork's humans no longer look at the CRM; agents populate it from meeting notes, emails, and team discussions, then track each company through the pipeline. Data room analysis, model rebuilding, and reference calls are automated, so the entry-level analyst role no longer exists at the firm. What remains for people is relationships and the decision itself. GPs should ask whether their firm's thinking is legible to a model, and LPs should ask what happens to firms where it is not.

Why It Matters

The consensus view is that AI is a productivity layer that every firm bolts onto existing workflows. Trivedi says bolting is the failure mode, and that the roughly 3,000 venture firms include many zombies that should die faster. On the other side are large franchises with legacy processes, junior staff pipelines, and back offices that cannot be rebuilt from scratch. Trivedi runs a five-person firm that markets itself as AI-native and admits building in public is part of its differentiation, so the argument is also a pitch. He concedes the returns proof is years away and that his own 2021 to early 2022 vintage was slow to respond to the ChatGPT moment.

Big Picture Drivers

Small firms can rebuild, large ones must retrofit: Footwork's five people and three or four aggressively run agents cost thousands of dollars a month. A firm with hundreds of staff and decades of process has no equivalent option.
The context layer is the asset: Agents moved from local laptops to an always-on cloud setup with a central Postgres database holding all firm context. Every pass, every internal debate, and every company Footwork wished it had backed feeds the agents that brief the next decision.
Five jobs plus firm building, and most get automated: Find, decide, win, help, exit, and firm building each have agents in progress. Trivedi expects fewer people per firm, a leaner back office, and analysts who become wielders of the tools rather than doers of the tasks.
An operations role that manages agents: Footwork hired a dedicated AI lead who builds agents, updates the context layer weekly, and feeds frontier learnings back into the investment worldview. Trivedi expects every firm, and most companies, to have this role.
Slope of learning as the founder filter: In a fast-moving world Footwork asks founders what changed in their thinking this week and watches for adjustment across meetings. Learning speed has become the trait weighted above all others.
Judgment is augmented today, a voter tomorrow: The agent has no vote yet. It surfaces the key questions and compares current spikes to past decisions and their outcomes. Trivedi expects it to become a decision maker as models and internal context improve.

By The Numbers

5: Full-time employees at Footwork, running 3 to 4 agents.
Thousands of dollars per month: Firmwide AI spend, well below tens of thousands.
~3,000: Venture firms at the industry's peak, many on their last fund, which Trivedi says is a healthy outcome.
3.5 years: Time since the start of the LLM era, the baseline for his expectation of how far models move in the next 3.5.
7: University endowments among Footwork's LPs after two funds, with fund one oversubscribed by March 2021 and closed April 13 after starting in January.

Key Trends to Watch

Solo capitalists returning at scale: Trivedi wrote about solo capitalists six years ago and now says a one-person firm with agents is easier to run than ever. Watch first-fund launches from single GPs with no analysts.
The analyst class shrinking across the industry: Footwork's conclusion that the analyst role is unnecessary will be tested by whether other firms hire fewer entry-level investors in the next two cycles.
Agents earning a vote: The observable marker is a firm that publicly assigns an agent formal input or a recorded position in investment committee.
Building in public as a sourcing channel: Footwork plans to publish more of its internal AI work to attract like-minded founders, while keeping a subset of portfolio companies unannounced at the founders' request.

Memorable Quotes

"An AI native firm is where the humans in the firm don't even look at the CRM because that's updated agentically." The test he uses to separate rebuilt firms from retrofitted ones.
"The entry-level analyst role, it doesn't feel like one that we need to have at Footwork." The first job to go, with the back office next.
"I just think it's healthy to have more death in our ecosystem." His view that firms unable to adapt should not survive 10 to 20 years.
"Exceptional companies deserve exceptions." The advice he gives new managers: bend ownership and stage rules for the handful of companies that make the industry's returns.

The Wrap

The thesis holds if Footwork's 2026 decisions, made with agents carrying the firm's full decision history, outperform its 2021 to 2022 decisions made without them, and if small rebuilt firms show better sourcing and selection than large franchises that layered tools onto old processes. It also holds if the analyst pipeline visibly contracts across venture and agents move from briefing to voting. It fails if firm-level context proves too thin to improve judgment, if large platforms with more data retrofit successfully, or if the operational savings never translate into returns. Trivedi says the answer arrives only when the cycle completes. The next 5 to 7 years of Footwork's fund two and fund three outcomes will deliver it.