A sourcing agent that sits in your team's Slack

2026-10-29 · Avi Agola, The Seattle Compute Company

One of our customers is a pre-seed VC firm. A lot of its early deal flow starts in places that aren't on a startup database yet: a thesis that points at a real product, a research grant that quietly funds a prototype. Finding those means reading, and reading doesn't scale across a small partnership's calendar. So the firm hired a sourcing agent.

What the agent actually did

1. 1.
   Finds new university research

   The agent reads recently published theses and grants and treats each as a possible company, noting who is behind it and what has been built so far.

2. 2.
   Discusses findings with the whole team in Slack

   It posts what it found in a channel with its own Slack handle. Partners reply in the thread, ask follow-up questions and push back on its read.

3. 3.
   Gets invited to Google Docs

   Someone shares the team's doc with the agent using the normal Share button. The agent then writes in it like a teammate, adding company write-ups and discussion notes.

Why it lives in Slack

The most useful part of this setup is not the research. It is that the agent is a participant in the team's existing conversation. It has its own Slack handle, like any other teammate. A partner tags it under a finding, asks "who else is working on this?", and the agent answers in the thread. A second partner sees the exchange without anyone forwarding anything.

That is different from a research tool that one person runs and then summarizes for everyone else. The reasoning stays in public view, which is where a partnership argues about whether a deal is any good.

Writing in the team's docs

Every firm keeps a memo or a pipeline in Google Docs. The agent gets access the way a person does: someone clicks Share and adds it. From there it writes in the doc like a teammate would, adding a section on a company or filling in the notes after a discussion.

Two things follow from that. You decide exactly which docs the agent can touch, because it only sees what's been shared with it. And the edits are the agent's own, attributed to it, so a partner reading the doc can tell what the agent wrote and what a human did.

What the record looks like

A sourcing agent reads a lot and writes into shared documents, so you want to know what it did. Smith traces every action to the agent that took it and the account it used. If a partner asks where a claim in a memo came from, the thread has the sources, and the log has the steps.

What to ask of a sourcing agent

A sourcing agent is only as good as the brief it works from. Three things help:

• Say what counts as a company. A thesis with a working prototype is not the same as a literature review. Tell the agent where your line is, in its role, once.
• Ask for the evidence, not just the verdict. The agent should link what it read, so a partner can check a claim in a click.
• Keep the discussion in the thread. When partners disagree with the agent's read, say so there. The thread becomes the record of why the firm passed or moved forward.

Writing the role well is most of the setup. Writing agent roles has a template.

What it does not do

The agent does the reading and the first-pass write-up. It doesn't decide which founders to call, and it doesn't replace the conversation a partner has with a founder. Those stay with people, which is where you want them.

Setting up the rest of the team

One sourcing agent is a start. The venture capital team guide walks through a fuller setup with sourcing, diligence and intro agents, and building an agent team explains how agents hand work to each other. If you want to see the same research-to-action pattern in another field, the AI recruiting agent page runs it for hiring.