Role playbook

AI for venture capital.

Venture is information processing with a judgment layer. AI compresses processing: more deals examined, faster and earlier. It does not improve the judgment that drives returns or create an edge every fund can buy.

This assumes you have the base. Start there if not

The job

What this role actually does.

Six things every week. Model releases do not change them; this page should still hold in three years.

01Sourcing 02Screening 03Diligence 04Memo 05Portfolio 06LP reporting

Most examples are vendor customer stories. Their self-reported, unaudited numbers show something was built and used—not a benchmark to expect.

Sourcing and pipeline

Research

The use case: Monitoring thesis areas continuously, and enriching inbound so a record arrives with context already attached.
Job to be done: When a company relevant to my thesis appears, I want to know early and with context, so I can build a relationship before the round is competitive.
Value to the business: More of the relevant market seen properly, and seen earlier.
How to evaluate it: Companies reviewed per partner, share of deals seen before the round opened, and how many sourced companies reached a first meeting. Everyone can monitor everything now: the edge is the filter and the relationship, not the feed.

What AI does here

AI monitors raises, departures, repositories, app stores and niche communities. Run a weekly thesis monitor and enrich inbound so partners open context, not just a company name and deck.

Workflow map

InputDefine the thesis, target stage and geography, then connect permitted public and inbound sources.
AI-assisted processMonitor signals, enrich company records and rank them against transparent thesis criteria.
Human checkpointInvestor reviews the highest-potential and deliberately samples rejected companies before outreach.
OutputA prioritised pipeline with context, rationale and a relationship-led next action.

What is still yours

Thesis and proprietary flow. Everyone can monitor everything; the filter differentiates. Knowing which sector matters and building a relationship before a competitive round is not a data problem.

Screening and first look

Extraction

The use case: Extracting the same fields from every inbound deck and screening against stated criteria.
Job to be done: When fifty decks arrive this week, I want each one read to the same standard, so the decision depends on the company rather than on who was on inbox duty.
Value to the business: A consistent, auditable first look, and faster, more respectful nos.
How to evaluate it: Time to first response, consistency of decisions on a re-run sample, and above all what a deliberate review of the rejects turns up. Every fund's returns come from a company that failed a criterion.

What AI does here

AI extracts the same fields from every inbound deck—stage, sector, geography, traction, raise and team—then screens against stated criteria. It also enables faster, more respectful nos.

Workflow map

InputReceive the deck, data room link and founder context alongside the fund’s stated criteria.
AI-assisted processExtract comparable fields, identify missing information and produce an explainable first-look summary.
Human checkpointInvestor reviews the source material, handles exceptions and chooses who receives a founder meeting.
OutputA consistent first-look record, prompt feedback and an auditable screening decision.

What is still yours

The exceptions. Every fund’s returns include companies that failed a criterion. Review rejects deliberately and give someone authority to overrule without a meeting.

Diligence

Synthesis

The use case: Interrogating the data room, market sources, references and technical material as one corpus, hunting for what does not add up.
Job to be done: When I am under a deadline, I want inconsistencies between documents surfaced, so the questions I take to the founder are the ones that matter.
Value to the business: Deeper diligence in the time available, and fewer surprises after the wire.
How to evaluate it: Issues found in diligence that would previously have surfaced post-investment, and time per deal. Every finding is traced to source before it enters a memo.

What AI does here

Put the data room, market reports, technical material, references and competitors in one corpus. Ask which deck claims lack support and where documents disagree.

Workflow map

InputCollect a permissioned data room, founder claims, market sources, call notes and diligence questions.
AI-assisted processSearch across documents, compare claims, summarise evidence and surface inconsistencies for follow-up.
Human checkpointDeal team traces each finding to source, runs reference calls and assesses the founder and risks.
OutputAn evidence log, open-question list and a diligence view ready for investment-committee debate.

What is still yours

Reference calls and the founder read. Whether they keep going after it stops working lives in tone, pause and what a former colleague does not say—not in a document.

The investment memo

Drafting

The use case: Drafting the factual sections, and generating the strongest possible bear case against your own recommendation.
Job to be done: When I take a deal to committee, I want the argument against it written down first, so I meet the partnership's questions in writing rather than in the room.
Value to the business: Better-tested decisions, and a memo that shows its reasoning.
How to evaluate it: Whether the risks named in the memo are the risks that actually materialise, reviewed at the next mark. Conviction cannot be generated: a committee can tell an assembled argument from a built one.

What AI does here

AI drafts factual memo sections, comparables, sizing and landscape. Use it adversarially: write the strongest bear case, then answer it before the partnership asks.

Workflow map

InputStart with validated diligence, comparable companies, a recommendation and the investment-committee template.
AI-assisted processDraft factual sections, compare alternatives and generate a strongest-possible bear case.
Human checkpointDeal lead checks every claim, names remaining uncertainty and writes the actual recommendation.
OutputA concise memo with evidence links, explicit risks and accountable ownership.

What is still yours

Conviction. A memo puts someone’s name against a probably wrong decision. Generated prose is fluent and unaccountable; committees know an assembled argument.

Portfolio support

Automation

The use case: Packaging what the fund knows into something founders can query, and matching needs against the network.
Job to be done: When a founder asks a question the fund has answered twenty times, I want them served immediately, so partner time goes to the conversations that need a partner.
Value to the business: Support that scales past the partner's calendar, and a reason for a founder to pick you.
How to evaluate it: Founder usage and satisfaction, and partner hours redirected to the hard conversations. Matching is mechanical; the introduction still needs someone who knows both sides.

What AI does here

Package fund knowledge founders repeatedly need: hiring and pricing material, templates and network matching. Helping portfolio companies adopt these tools is useful service.

Workflow map

InputCurate approved playbooks, operator knowledge, portfolio needs and introducer permissions.
AI-assisted processAnswer repeatable questions, match needs to approved resources and draft context-rich introductions.
Human checkpointPartner confirms the advice fits the company and personally handles sensitive introductions or crises.
OutputFaster founder self-service and a warm, accountable intervention where judgment is needed.

What is still yours

Show up when it goes badly: the down round, co-founder split or overdue firing. Value-add content is table stakes.

LP reporting and fund communications

Drafting

The use case: Standardising portfolio updates, normalising the numbers, and drafting the quarterly report.
Job to be done: When reporting is due, I want assembly handled from standardised inputs, so the team spends its time on what the numbers mean.
Value to the business: Analyst weeks returned each quarter, and reports that go out on time.
How to evaluate it: Cycle time per report, restatement rate, and LP questions received after the fact. Valuations, disclosures and the framing of a bad quarter are judgment calls the GP signs.

What AI does here

Quarterly reporting is assembly from known inputs. Standardise updates, extract consistently and generate the draft; apply the same approach to fundraising materials.

Workflow map

InputCollect standardised portfolio updates, fund accounting, valuations, prior letters and disclosure requirements.
AI-assisted processExtract and normalise KPI updates, reconcile variances and draft a report using approved language.
Human checkpointFinance and partners validate figures, valuation judgments, confidentiality and the narrative on material changes.
OutputAn accurate LP report with a clear, candid explanation of performance and next steps.

What is still yours

What you disclose and how you frame a bad quarter. LP relationships depend on hearing the truth early, not eventually.

Start this week

Extract the same ten fields from your last fifty inbound decks and look at what your screening has quietly been biased toward.
On your next live deal, ask for the strongest possible bear case before you start writing the memo.
Set up one weekly monitor on your thesis area so that sourcing stops depending on who remembers to go looking.

Where it fails in this role

Picking. If these tools were predictive, returns would have compressed by now and they have not. The alpha is access and judgment, not processing speed.
Founder assessment. Nothing in a document tells you whether they will still be there in year four.
Proprietary insight. Public data reaches every fund simultaneously, so an edge built on it is not an edge.

Other playbooks

All nine playbooks Product management Sales Marketing Operations Consulting Forward deployed engineer Founder / entrepreneur AI product manager

Now do it somewhere real.

A playbook is a map. The Industry Fellowship is the terrain: a term spent talking to professionals who are implementing AI in one industry, working out where it actually creates value, and building a working prototype against what you find. Most fellows start as beginners.

Explore the Fellowship Apply for the Class of 2028