Pipeline Discipline
Click the buttons. Every stage transition is a decision — and the AI stack only works on deals the CRM actually tracks.
The diligence stack — the memo generator, the competitor finder, the contact surfacer, the firm's internal intelligence — only works on deals the CRM knows about. If a company has had four calls, a data room review, and a partner meeting but its stage still reads “Targeted” (or worse, sits in “Universe”), the AI has no idea this deal is live. Future teammates won't either. Clicking the buttons isn't admin — it's the input layer the whole system runs on.
Every stage is a decision
An investment decision is not a single moment. It is a series of smaller decisions accumulated over time, each one a gate a company either passes through or does not. First call to team discussion is a decision. Team discussion to initial diligence is a decision. By the time a firm is deep in active diligence, they have already made five or six conscious choices to keep going. That accumulated conviction is not accidental — it is the product of a disciplined process.
Failing to make a decision at any stage is not neutral. It is, in effect, a bad decision — one made by default rather than by judgment. A company that lingers without a stage change is quietly consuming cognitive bandwidth, creating a false impression of pipeline depth, and obscuring where the process has actually broken down.
The exact stages, gates, and checks will vary firm to firm — there is no universal template. What matters is that the firm has written them down, agrees on what each one means, and uses them the same way every time. How you assemble these into a repeatable process is up to you. Two reference checklists below: the first is the investment process — the work you do on a deal to decide whether to invest. The second is confirmatory due diligence — the verification work that happens once you've decided you want in, to make sure everything checks out before funds move.
Reference: Investment Process Checklist
- Add company to the CRM
- Add company to the deal pipeline list
- Generate Pre-Meeting Brief (PMB) by flagging the company in the PMB workflow
- Send intro email to founder — compose manually or use the auto-drafted intro on the company record
- Take intro call — founder-led pitch covering technology, business model, competitive positioning, and capital needs
- Log call notes, learnings, and proposed next steps in the CRM
- Surface company in internal and broader team calls
- Pass — mark CRM status as Pass
- Reconnect — set status to Reconnect, define a reconnect trigger (e.g., “raising in 3 months”), and create a task to fire on that date
- Continue — proceed to Stage 5
- Create a dedicated file-drive folder for the company
- Confirm the key diligence page auto-creates alongside the folder
- Request data room and supporting documents from founder
- Upload data room contents (accessible to team and AI tooling)
- Run an AI-powered data room inventory to flag missing datapoints and request them
- Schedule a round / deal-terms call when the round comes into focus — clarify valuation, leads, structure, timeline, and own role
- Run deep research queries against key open questions from the intro call
- Populate the diligence page with intro notes, open questions, and initial views
- Synthesize findings to inform the go-live decision
- Update CRM status to Live
- Schedule follow-up call with broader team; deeper Q&A on key open questions
- Run key diligence contacts finder across data providers
- Populate diligence page with prioritized expert/reference list and structured sections for market, customers, technology, team, and risks
- Begin scheduling technical-expert and reference / investor-validation calls, highest priority first
- Generate the two-pager Investment Committee Brief and circulate
- Pass — move to Reconnect with defined trigger and task
- All-In — proceed to Stage 7
- On-site / in-person technical deep dive — see the product, meet the full team, go deep on engineering and supply chain
- Run remaining technical-expert and reference / investor-validation calls
- Financial model deep-dive call with CFO or finance lead — unit economics, revenue ramp, capital structure, key assumptions
- Log all calls and transcripts to the file drive
- Collaboratively build out the diligence page — call notes, answers to key questions, evolving thinking
- Present in team / partner meeting
- Capture feedback from senior partners and external advisors
- Build financial model: revenue build, cost structure, capital plan, dilution scenarios
- Benchmark against comparable private and public companies (growth, margins, capital efficiency, multiples)
- Run returns analysis: base, bull, and bear MOIC / IRR at entry valuation
- Stress test against capital intensity, time-to-revenue, and exit assumptions
- Upload model and benchmarking deck to the file drive
- Generate first-draft Investment Committee Memo from the diligence page
- Layer in accumulated data, call transcripts, model output, benchmarking, and the team's thinking
- Produce v1 ICM for broad team distribution
- Host IC Round 1 — surface and nail down outstanding questions
- Document open items and assign owners
- Kick off the confirmatory due diligence checklist
- Address all open items from Round 1
- Update model, benchmarking, and ICM with new findings
- Recirculate updated ICM ahead of Round 2
- Host IC Round 2 — final invest / no-invest decision
- Invest — move to Portfolio, kick off close, finish the confirmatory diligence checklist
- Not — move to Reconnect with defined trigger and task
Reference: Confirmatory Due Diligence Checklist
Use this after market, product, technology, and competitive diligence are substantially complete. The objective is to verify the facts that can reasonably harm the investment: ownership, economics, cash, revenue, debt, liabilities, IP / control rights, governance, regulatory exposure, and integrity. The confirmatory standard is not exhaustive legal perfection — it is whether each material risk is cured, priced, protected, or consciously accepted before funds are wired.
- Start from materiality: spend legal and advisory dollars on issues that can impair ownership, exit-ability, future financing, customer eligibility, or trust in management.
- Evidence: cash, revenue, debt, cap table, IP ownership should tie to independent records or direct confirmations.
- Treat fraud signals as separate from ordinary business risk (refusal to confirm, contradictory documents, undisclosed side letters, unexplained related-party activity require escalation).
- P0 — Deal breaker / IC re-approval. Fraud, intentional misstatement, invalid security issuance, missing core IP ownership, sanctions / export / gov't violation, material undisclosed liability, or integrity issue. Stop or return to IC.
- P1 — Closing condition. Curable before close. Make a hard closing condition; verify completion before wire.
- P2 — Structure / price / rights. Address through valuation, terms, indemnity / escrow, covenants, board rights, or reserves.
- P3 — Post-close governance. Real but manageable. Add to 100-day plan with a named owner.
- Cash and bank accounts — direct bank confirmations, statements, reconciliations, GL tie-out, closing cash certificate.
- Revenue — top customer contracts, invoice and cash-receipt samples, AR aging, customer confirmations, side-letter certification.
- Debt, liens, liabilities — debt schedule, lender confirmations, payoff letters, UCC / lien searches, AP / accrual support, tax filings.
- Cap table and security — counsel cap-table reconciliation, charter, stock ledger, SAFEs / notes / warrants, board / stockholder approvals, side letters.
- IP ownership and control — founder / employee / contractor assignments, patent and trademark schedule, open-source scan, university or customer-funded development review.
- Management integrity — background checks, reference calls, civil / criminal / lien / judgment searches, prior investor and board feedback.
- Material contracts and consents — top customer / vendor review, required consents, termination / assignment / change-of-control / MFN / SLA review.
- Regulatory and compliance — sanctions / export / CFIUS / gov't-contracting / privacy / cyber screen; specialist memo when triggered.
- Closing controls — officer certificate, good standing, board / stockholder consents, disclosure schedules, funds flow, verified wire instructions.
- Process integrity — freeze final data room, track open items in a severity-ranked issue log, confirm no undisclosed side letters or related-party transactions.
- Financial — review monthly financials / TB / GL, cash proof against bank statements, sample top customers and period-end invoices, identify add-backs and debt-like items.
- Corporate / cap table — reconcile fully diluted cap table; review SAFEs, notes, warrants, options, RSUs, side letters; confirm protective provisions and required consents; review 409A.
- Governance / integrity — confirm board / observer / voting / information rights, review minutes for material actions, identify related-party transactions and insider economics, complete background checks on key executives.
- IP / data rights — confirm assignments across founders / employees / contractors / universities; review patents, trademarks, repos, models, datasets; assess open-source and customer-funded dev rights; confirm no IP liens or threatened claims.
- Contracts — review top customer and critical vendor contracts; check termination, assignment, change-of-control, MFN, exclusivity, SLA, refunds, indemnities, liability caps, minimum commitments; confirm all side letters and concessions.
- Litigation, tax, insurance, employment — confirm tax filings (payroll, sales / use, R&D credits, 409A, transfer pricing); review executive employment, severance, change-of-control, commission, immigration; confirm D&O, cyber, E&O, product liability, EPLI, workers' comp.
- Regulatory triggers — sanctions, anti-bribery, export-control, compliance screen; EHS / product liability when hardware, energy, materials, food, or manufacturing risk exists.
- Closing documents — confirm security economics match IC approval; review SPA, charter, IRA, voting, ROFR / co-sale, side letters; finalize disclosure schedules and bring-down reps.
- Independent confirmations — cash with banks, debt with lenders, top customer contract and payment / side-letter status, auditor relationship (if audited), insurance coverage where material. Hard stops: refusal to permit direct confirmation, customer or lender contradicts management, only screenshots or internal evidence provided.
- Tie-outs — cap table to charter, stock ledger, approvals, financing docs; payroll / headcount to payroll provider and HR roster. Hard stops: numbers don't tie across systems, definitions change late, large unexplained period-end transactions.
- Pattern tests — manual journal entries near period-end, credit memos / refunds after quarter-end, round-dollar or unusual invoices, related-party vendors / customers, unusual cash transfers or aging. Hard stops: recurring quarter-end cleanups, past-due customers counted in ARR, founder-affiliated transactions not pre-disclosed.
- Human intelligence — founder / executive background checks, prior investor and board member calls.
- Confirmatory diligence memo — scope completed, key evidence, independent confirmations, open issues, severity, decisions, required close conditions, post-close actions. 3–5 pages, exception-based, signed off by deal lead.
- Issue register — workstream, description, severity, owner, economic / legal impact, resolution path, deadline, final status. No P0 / P1 unresolved at funding; P2 / P3 each have an owner and mitigation.
- Fraud verification certificate — cash, debt, revenue, customer, cap table, background checks, lien / litigation, related-party, and wire checks completed before funds flow; exceptions escalated.
Why this matters more with AI in the loop
Pre AI, sloppy pipeline hygiene was a coordination problem. Post AI, it is a data problem that compounds across every system the firm runs:
- The memo generator pulls from active diligence deals. If the stage is wrong, the deal doesn't show up — or worse, the wrong deals do.
- Internal intelligence (RAG over the corpus) uses stage to weight relevance. A “Targeted” company looks like cold pipeline; a “Passed” one becomes a benchmark. Mislabeled deals poison both signals.
- Workflows and orchestration trigger on stage changes — scheduling reference calls, pulling enrichment, alerting the partner. No stage change means no automation.
- New team members read the pipeline to learn how the firm thinks. An untended pipeline teaches them the wrong lessons.
Define the stages, then defend them
Well defined stages force the question at every gate. Each stage has a clear meaning and a clear action associated with it. Moving a company forward requires a judgment. Keeping it in place implicitly requires one too. The minimum viable definition for each stage:
- What it means — the specific event that puts a company in this stage (e.g., “first call complete, partner interested”).
- What action is owed — the next concrete step (request the data room, schedule a team review, draft a memo).
- Who owns it — one named person, never a team.
- The exit criteria — what has to be true to advance, pause, or pass.
Passing is a stage change too
The most under-used button in any pipeline is “Pass.” Firms hesitate because a pass feels final, or because the company might come back around. Neither is a reason to leave a deal in limbo. A logged pass is a valuable data point that compounds over time into institutional knowledge about what the firm does and doesn't believe in. An unlogged non-pass is just noise.
The accountability layer
When stages are clearly defined, the team always knows who owns what and where things stand. Nobody steps on each other's toes reaching out to the same company. Nobody has to ask whether a first call has happened. The pipeline tells the story of every company's journey through the firm's evaluation process, and that story is readable at a glance — by humans and by the AI layer alike.
The bar
Every company that's had a real interaction has a stage that matches reality. Every stage change happens within 24 hours of the event that triggered it. Every deal has a named owner. Passes are logged, not ghosted. The pipeline is not a filing system — it is a record of judgment, exercised consistently, over time. That is what separates a disciplined investment process from a reactive one, and it is what gives the AI stack something real to work with.