How investors run the public-data layer of diligence in an afternoon using AI

The pre-IC screen burns analyst-days on tab-hopping, skips the sentiment check, and quietly averages three databases into one confident-sounding number. This guide walks a workflow that does the public-data layer properly: registry verification, the company as it presents itself, what the public actually says, and red flags — every claim traceable to a source, every estimate labeled as one — capped with a pre-mortem that turns risks into watchable signals.

The screen starts with one message:

“Run a due diligence brief on [company]. Context: investment screen. Pay extra attention to customer-side sentiment.”

Why this isn’t just a ChatGPT prompt

  • It reaches registries, not just search results. Legal entities across 309 jurisdictions: status, incorporation date, LEI, registered address — and same-name entities abroad, which map subsidiaries before the data room does.

  • It presents disagreement instead of averaging it. When three sources give three customer counts, the brief shows all three with sources named. A generic prompt blends them into one confident wrong number.

  • Estimates are labeled, permanently. A third-party revenue estimate is carried as “estimate, single source” and never restated as fact. At IC, that labeling discipline is the whole game.

  • The sentiment leg reads what nobody markets. Community threads — customers and counterparties talking when nobody’s selling to them — with proportionality rules built in: one bad thread is not a pattern, and no-data is reported neutrally.

  • Findings, not verdicts. No scores. The workflow is explicitly forbidden from rating companies. It brings evidence and open questions; the judgment stays yours — which is exactly what an IC will demand.

  • The pre-mortem produces signals, not adjectives. Two named failure narratives with early-warning indicators a deal team can put on a dashboard. “Risks: competition” doesn’t survive it.

The steps

  1. 1.Step 1 — run the diligence workflowPro

    Give it the company name, the context, and your specific concern. It drives its own research: registry search and full record, LinkedIn profile and activity, grounded web coverage (including the complaints sweep), and community sentiment.

    Say this

    “Run the Company due diligence brief workflow on [company]. Context: investment screen. Pay extra attention to [your concern].”

  2. 2.Step 2 — read the cross-references, not just the sections

    Registry vs. pitch. Customer counts across sources. Sentiment vs. positioning. The synthesis step checks sources against each other — and where they disagree, the disagreement itself is the finding, deferred to the data room for resolution.

  3. 3.Step 3 — run the pre-mortemPro

    Feed the brief to the Pre-Mortem Generator (Kahneman): assume the investment failed, work backwards. Out: two specific failure narratives — internal and external — with credible risk factors and their early-warning signals.

    Say this

    “Run the Pre-Mortem Generator on this investment thesis, using the diligence brief.”

  4. 4.Step 4 — take it into the room

    The brief’s next-steps section becomes your management-questions list; the discrepancies go to the data room and reference calls. The IC memo is still yours — now it starts from evidence.

What it looked like when we ran it

We ran the full chain on a real fintech (anonymized; not fundraising — the deal framing was our test fiction) and published the brief. Three moments from that run:

The registry leg mapped the legal footprint in minutes

Active UK entity since 2012, LEI on file — plus two more active same-name entities in Peru and Japan. Week-two data-room work, done before the first call.

The sources disagreed, and the brief said so

Customer count: 10,000+ said the company, 6,000+ said one database, 4,000+ said another — presented side by side with a note to ask management for the audited number and its definition.

Sentiment became a management question, not a verdict

A cluster of small-seller complaints (suspensions, 180-day payout holds) cross-referenced against crawl data showing 82% of the customer base has ten or fewer employees. The segment driving signups is the segment risk-ops squeezes — exactly the question the deal team takes into the data room.

The associate and fund in the sample are invented; the target company is real (anonymized) and every fact is public record — registry, LinkedIn, press databases, public forum threads.

Common questions

Does this replace real diligence?

No — it’s the public-data layer, done in an afternoon instead of an analyst-week. Your data room, expert calls, and model still do what only they can; they just start from sourced claims and pre-written management questions.

How does it handle conflicting numbers?

It presents the disagreement with each source named, instead of averaging. Estimates are labeled as estimates with their source class. Resolution goes to the data room.

What about companies with thin public footprints?

Gaps are reported neutrally — “no registry match” or “no community discussion” is stated as a fact with context, never spun as suspicious or silently skipped.

Can the same chain vet vendors, not just investments?

Yes — the workflow is built for investors, procurement teams, and anyone vetting a company before signing. Same brief, different context setting.

Run this guide on the next deal that moves to IC

The due diligence workflow, the registry and research tools, and the Pre-Mortem Generator are in Amplifiers Pro.

Try Amplifiers Pro today

Works inside Claude · setup takes minutes · cancel anytime