Skip to main content
Book a call
Workshop

Cohort 18 lecture. For VC firms.

How VC firms are using AI.

Venture is a knowledge-work business. Knowledge work is what AI is eating first. Here's what that actually looks like inside a fund in 2026, and what it still can't touch.

01 · The shift

From asking AI questions to delegating real work.

Most professionals are using AI in a way that's already obsolete. There are three modes, and the gap between them is the gap between fast funds and slow ones.

01

Chatbot user.

Open ChatGPT in a tab. Ask it questions. Copy the answer back into your doc. AI is a separate tool you visit.

Where most people are
02

Copilot user.

AI lives inside the tools you use. It autocompletes emails, summarises docs, drafts memos. Helpful, but you're still doing the work.

Where good analysts are
03

Agent operator.

You delegate whole tasks. The agent reads your inbox, pulls the deck, runs research, drafts the memo, files it in Notion. You review.

Where the leverage is

The fund of 2026 has analysts who operate AI.

02 · A real example

Building a competitive landscape.

Partner forwards you a deck on Tuesday morning. "Get me up to speed on this market by EOD." Here's what the next four hours look like for most analysts.

  1. 01 Read the deck. Pull out the company, product, market.
  2. 02 Open PitchBook, Crunchbase, Tracxn. Search the category.
  3. 03 Copy 30 competitors into a Google Sheet.
  4. 04 Open every competitor's site. Note product, pricing, positioning.
  5. 05 LinkedIn each founding team. Headcount, pedigree, hiring trends.
  6. 06 Search news, podcasts, Reddit, X for signal on each one.
  7. 07 Paste it into ChatGPT to summarise.
  8. 08 Build the comparison table. Pick the 5 closest competitors.
  9. 09 Write the summary memo. Send to partner at 6pm.

4 to 8 hours. Six tools. And you're the glue between every step.

03 · The insight

Every analyst action is a tool call.

Every action you take in a tool, an agent can take too. Buttons, searches, exports, copies: they're all just functions underneath.

What an analyst does = What an agent calls
Parse a pitch deck PDF = pdf.extract()
Search PitchBook for category = pitchbook.search()
Read 20 company websites = web.fetch()
Look up founders on LinkedIn = linkedin.lookup()
Write a memo in Notion = notion.create()
Post the summary in Slack = slack.postMessage()

Every step in that four-hour workflow is a tool call. An agent can chain them all together and do the whole job in one shot.

04 · The agent way

Same task, delegated.

Here's that same competitive landscape task as a single delegation to an agent that has access to your tools.

Prompt

"Here's a deck for <Company>. Build me a competitive landscape with 5 closest comps, what they sell, pricing, headcount, last round, founder pedigree. Drop it in Notion as a memo."

Working

Agent reads the deck, runs the searches, opens the websites, scrapes LinkedIn, drafts the memo, files it.

Result
  • Deck parsed, category extracted.
  • 30 comps surfaced, 5 ranked by closeness.
  • Pricing, team, funding pulled per comp.
  • Memo drafted and filed in Notion.

You speak. It acts. Four hours of admin work becomes thirty minutes of strategic review.

05 · The map

Where AI lives inside a fund.

Walk through any VC's week and the same pattern shows up across multiple workflows. This is the map.

01 Sourcing

Finding deals before everyone else.

Today

Manually scanning Twitter, newsletters, LinkedIn. Asking around. Hoping the right founder finds you.

With agents

Agents monitor your inbound, score signals across LinkedIn, X, Substack, and news, mine your own network for warm intros, and surface a ranked weekly list with reasoning.

02 Diligence

Going deep, fast.

Today

Days of research per company. Reading reports, calling experts, building models. The bottleneck on how many deals you can seriously evaluate.

With agents

Custom research agents gather and synthesise data from 100+ sources per query. Domain-trained. Source-cited. Ready in minutes instead of days.

03 Memos & IC

From research to recommendation.

Today

Dozens of hours per memo. Analyst writes draft, partner edits, back and forth. The bottleneck between "we like this deal" and "we filed the IC memo."

With agents

Agents trained on your fund's thesis and memo format. Pull the research, structure to your template, cite sources. Analyst reviews and refines.

04 Portfolio support

Real-time, not quarterly.

Today

50 portfolio companies, 50 different reporting formats, quarterly chase-down. By the time you see a problem, it's been three months.

With agents

Agents normalise KPIs across messy formats. Anomaly alerts when something drifts. Partners walk into board meetings with current data, not stale reports.

05 Fund operations

LP comms, ops, the boring stuff.

Today

Quarterly LP letters. Data room maintenance. Comp sweeps. Hours of work that doesn't make a single deal happen.

With agents

Draft LP updates from portfolio data. Auto-update data rooms. Run comp benchmarks on a schedule. The ops layer runs itself.

06 · The limit

The relationships are still the job.

Most of what AI replaces in a venture firm is the work that wasn't actually the job to begin with.

What AI absorbs.

  • Research and synthesis.
  • Data gathering and normalising.
  • Memo drafting from your thesis.
  • Triage, sorting, scoring.
  • Report generation, LP updates.
  • Pipeline ops, data room maintenance.

What stays yours.

  • Building real conviction in a founder.
  • Making your case during an IC.
  • Helping founders through a bad day.
  • Knowing when to break your own thesis.
  • The texture of an in-person introduction.
  • Earning the trust of an LP over years.

Use agents around the relationships and deals. Automate the work that takes time away from thinking strategically about an investment and building relationships with founders.

07 · Your move

How to actually start.

The funds making AI work are the ones who picked one workflow and got serious about it.

  1. 01

    Pick the workflow that bleeds time.

    Where do you lose four hours a week to copy-paste? That's your candidate. Sourcing triage, comp sweeps, deck summaries, LP updates: pick the one that hurts most.

  2. 02

    Map the steps. Separate judgment from mechanics.

    Write out every click. The mechanical steps go to the agent. The judgment calls (what to elevate, who to back) stay with you. Most workflows are 80% mechanical.

  3. 03

    Build the agentic version in days, not months.

    Skip the perfect platform. Tools like Claude, n8n, and a few hours with someone who can wire APIs together gets you 80% there. Iterate from there.

  4. 04

    Watch it work. Fix what it gets wrong.

    First version will miss things. That's data. Every correction tightens the agent. Within a month it's doing the job better than the analyst who used to do it.

The moat now is knowing what to build.

Talk to us

Want this workshop for your fund?

We run this lecture as a private session for partner teams and analyst cohorts. 30 minutes is enough to know whether it's a fit.

Book a call Or write to us at