Your AI Investing Team: What the Agents Do All Week

An AI investing team is a set of agents that run your fund's operations on a continuous cadence: finding companies, prepping diligence, keeping your deal context current, and turning portfolio updates into intelligence. On GPAgent, that team runs all week on your behalf and leaves a record of everything it did. You stay in the chair for the decisions that matter: which founders to meet, what to invest in, and what to say.

This is the part of GPAgent that is easy to undersell. Underneath, yes, there is a durable fund record and an API. But a record with an API is just a filing cabinet. The product is the team that works the cabinet for you every day. If you run a fund alone, you have been that team. This is what it looks like to hand the operations to agents and keep the judgment.

GPAgent runs five investing roles, each owning one job:

1. Sourcing Analyst finds and qualifies companies, on a weekly sweep.
2. Investment Analyst scores the best ones and preps them for a founder conversation, daily.
3. Relationship Manager reconciles your calendar and meetings into your deal cards, nightly.
4. Portfolio Analyst turns portfolio updates into metrics, risks, and LP-ready intelligence, weekly.
5. GP Brand Strategist turns verified fund work into public narrative.

Not a database with an API

Most "fund OS" tools are a place to put things. You still do the work; they just store the result. That is a better filing cabinet, not a team.

An AI investing team inverts that. The agents do the work and the record is where they write it down. The Sourcing Analyst does not wait for you to paste a batch into a table; it sweeps your sources and fills the pipeline. The Relationship Manager does not wait for you to log a meeting; it reads your calendar and updates the deal card. The record still matters enormously, because it is what makes the team auditable. But the record is the floor. The team is the product.

That distinction is the whole reason GPAgent runs on a system of record with one writer per job. When agents are doing continuous work on your fund, you need to know exactly who changed what and why. A team without that is chaos. A record without the team is a nicer spreadsheet.

A week in the life of the team

Here is what "runs continuously" actually means. The cadence is the point: each role works on the rhythm its job needs, so the operations never go cold while you are in meetings or asleep.

Role	Runs	What it does for you
Sourcing Analyst	Weekly sweep	Pulls new companies from YC and your configured sources, scores them against your thesis, writes qualified deals, sends a digest
Investment Analyst	Daily	Reviews the strongest sourced companies, adds reasoning, promotes the best toward a founder conversation, holds the rest with a reason
Relationship Manager	Nightly	Reconciles your calendar and meetings into deal cards, moves met companies to a decision, attaches recording links
Portfolio Analyst	Weekly	Reads portfolio updates, extracts metrics and risks, flags where you can help, drafts LP-ready intelligence

Sunday: the Sourcing Analyst sweeps

While you are not working, the Sourcing Analyst is. It pulls the latest companies from the YC explorer and your other configured sources, scores each against your investing lens, writes the qualified ones into your pipeline as structured deals, and emails you a digest with the reasons and the anomalies. You open Monday to a triaged queue, not a pile of links.

Every day: the Investment Analyst preps the queue

Each day the Investment Analyst works through the strongest sourced companies, writes reasoning you can actually trust, and promotes the best ones toward a founder conversation. It is capacity-aware on purpose: when your founder-conversation queue is full, it keeps analyzing but holds promotions and tells you why, so your attention stays scarce and intentional instead of flooded.

Every night: the Relationship Manager keeps the deals honest

This is the job that quietly rots when you are busy. The Relationship Manager runs nightly against your calendar, matches events to companies by attendee domain and founder name, and updates the deal card to match reality: this founder conversation is scheduled, that one already happened and now needs a decision. When a recording link exists, it attaches it to the right card. No more deal board that lies to you because you forgot to update it after a call.

Weekly: the Portfolio Analyst compounds what you already own

After you invest, the Portfolio Analyst reads incoming portfolio updates, pulls out metrics and risks, notes where you could actually help a company, and turns the raw progress into LP-ready intelligence. The investor update you used to dread writing starts from real, current data instead of a blank page.

What stays yours

The team runs the operations. It does not run the fund. That line is deliberate, and we will not cross it.

The Sourcing Analyst can flag a company; it does not email the founder. The Investment Analyst can build the case and the open questions; it does not decide to write the check. The Portfolio Analyst can draft the LP update; it does not hit send to your LPs. Founder conversations, investment decisions, and outreach stay with you, because those are relationship and judgment calls, and a fund that automates them away is automating away the reason its LPs showed up.

What you get back is the hours around those decisions. The triage, the prep, the reconciliation, the first draft: that is what the agents take.

How you know the work actually happened

A team you cannot inspect is worse than no team. So every run writes to an activity ledger we call RunActivity: what ran, which role ran it, what it changed, what it checked, whether delivery succeeded, and what is safe to do next.

This matters more than it sounds. A scheduler reporting "done" is not proof of anything; it is a green light with nothing behind it. The ledger is the difference. It can tell you the Portfolio Analyst read eleven updates, wrote four metric snapshots, found one data gap, and flagged it, or that a run failed halfway and is safe to retry. You can audit the team the way you would audit an associate, which is the only way you should ever trust software to write to your fund.

Hosted by default, open by design

GPAgent owns the clock and the hosted execution path for production runs. You connect Gmail, Calendar, Fathom, Drive, and the rest of your fund sources in GPAgent, authorize each role, and watch the team run with a ledger underneath it.

The API still matters. OpenClaw, Claude Code, MCP clients, cron jobs, or your own scripts can connect to the same record when you have a specialized workflow. The trigger is replaceable. Your fund's team and its history are not.

The GP Brand Strategist: narrating the work

The fifth role is the GP Brand Strategist. It turns real, verified fund work into public narrative: market observations, founder lessons, portfolio themes. It runs on the same ledger as the rest of the team, which is exactly what makes it trustworthy. It narrates work that actually happened, and it does not invent a story or leak a private one. The other four agents earn the fund its results; this one turns those results into a public track record that compounds your deal flow and your standing with LPs.

FAQ

What is an AI investing team?

It is a set of AI agents that each own one fund job and run on a continuous cadence: sourcing, investment analysis, relationship management, and portfolio monitoring. On GPAgent the team runs your operations week to week and records everything it does, while you keep the founder conversations and investment decisions.

Does the AI investing team make investment decisions?

No. The agents prepare the work: triage, scoring, diligence prep, reconciliation, and first drafts. Investment decisions, founder outreach, and LP communications stay with you. The team buys you the hours around the decisions, not the decisions themselves.

How is this different from a fund CRM or "fund OS"?

A CRM is a place to store what you did. An AI investing team does the work and writes the result into the record for you. The system of record still matters, because it is what makes the agents auditable, but it is the floor. The continuously running team is the product.

How do I know the agents actually did the work?

Every run writes to the RunActivity ledger: what ran, what changed, what was checked, and whether it succeeded. A green scheduler is not proof; the ledger is. You can inspect each role's runs the way you would review an associate's work.

Can I use my own agent runtime with GPAgent?

Yes, as a specialized extension or integration. GPAgent runs the hosted team, and OpenClaw, Claude Code, MCP clients, cron jobs, or your own scripts can write through one API to one fund record.

What does the GP Brand Strategist do?

It drafts public narrative from verified fund activity in the ledger: market observations, founder lessons, and portfolio themes you can publish. Because it is grounded in real recorded work, it does not invent examples or expose private fund data. It turns the team's results into a public track record without you starting from a blank page.