How Marc Palet built a one-person VC back office with AI

At 26, Marc Palet is the sole Asia-Pacific employee of venture capital firm OMVC, based in Singapore. Rather than hire an intern or junior analyst to handle operational tasks, he has stitched together a suite of artificial intelligence tools that costs him roughly $120 a month. His subscriptions include chatbots, transcription services and coding assistants, and he has built custom agents that automate everything from meeting notes to quarterly investor reports.

Palet’s AI journey began in 2023 with ChatGPT. The real turning point, however, came when he adopted coding tools like Claude Code, Cursor and Replit. His first major build was a workflow that automatically summarizes recorded calls with startup founders. Instead of sitting through replays, he gets a structured memo covering the company’s problem, solution and market — exactly the information he needs for investment decisions.

He later created a personal knowledge repository using NotebookLM, feeding it every interesting article or post he read. That allowed him to query his own reading list when writing blog posts. More recently, he built an agent that ingests information from Gmail, Slack, his call transcripts, newsletters and YouTube podcasts, then keeps a living, organized page for each portfolio company. The agent does not just paste snippets; it rewrites sections so the whole story stays coherent. Another workflow compiles quarterly limited-partner reports by summarising portfolio activity automatically.

Palet says the payoff is tangible. His LinkedIn article output has jumped from one deep-dive every two months to one every three weeks, driving more deal flow. And because he spends fewer hours on routine paperwork, he has more time for face-to-face meetings. His conclusion: “AI plus me is more powerful than me plus an intern.”

Advertisement

Inside the economics of AI agents versus a junior hire

The real cost difference: AI subscriptions vs. an intern

Palet’s $120-a-month AI stack is a fraction of the cost of even a part-time intern in Singapore. A full-time junior analyst would also require onboarding, supervision and time spent correcting mistakes — a cycle he says can take a month of training for a three-month stint. The AI tools, by contrast, require upfront setup but then follow his exact specifications without variation. That does not mean AI is a drop-in replacement for human judgment; Palet still makes the investment calls and maintains relationships. The agents simply absorb the repetitive, time-consuming tasks that would otherwise fall to an entry-level hire.

Where agentic AI fits in a small VC’s workflow

Palet’s case illustrates how far off-the-shelf AI has come for process automation. Three years ago, a similar setup would have required a developer and significant maintenance. Today, coding assistants let a non-developer “vibe-code” custom agents that parse unstructured data from multiple sources and rewrite it coherently — a task that even many enterprise tools struggle to get right. For small funds with no dedicated back-office staff, the ability to auto-generate LP reports or audit material can be a genuine competitive advantage.

The limitations Palet isn’t facing yet

So far, Palet’s workflows have functioned without major errors, but the arrangement leans heavily on the reliability of third-party services and on the quality of the training data his agents ingest. If a transcription API misses a critical figure or an agent misinterprets a Slack thread, the error could cascade into a quarterly report that misstates a portfolio company’s financials. For now, Palet’s oversight catches those risks, but as the volume of automated information grows, the risk of unnoticed mistakes does too. That is the trade-off any lean team must weigh before swapping a person for a pipeline of prompts.

What lean teams can steal from Palet’s $120-a-month experiment

  • Audit the repeatable before you automate. Palet identified four specific drains on his time — meeting notes, deal-flow research, LP reporting and LinkedIn content. Map your own team’s tedious, rule-based tasks before choosing tools; the AI stack should mirror what you actually need to offload, not a generic wish list.
  • Start with a multi-tool mix, not a single platform. Palet combines different AI services for different jobs — chatbots for drafting, transcription APIs for calls, coding assistants for building agents. No single product handled everything, so his cost stayed flexible and he could swap components that underperformed.
  • Build a self-updating knowledge base early. The portfolio-company pages that rewrite themselves are the most sophisticated piece of Palet’s stack. Even a simpler version — a shared notebook that auto-summarizes meeting notes and email updates — can end the scramble before every board meeting or investor update.
  • Measure output, not just time saved. Palet’s increased publishing frequency directly boosted his deal flow. Pick a leading indicator that matters to your business — proposals sent, client calls scheduled, bugs fixed — and watch whether AI shifts it, rather than only counting hours that didn’t sit in front of a spreadsheet.