Project Atlas

A partner wants an initial screen by Friday, and a 62-page CIM has just landed. This page follows one fictional target, Project Atlas, from that moment to a drafted section of an investment committee memo. Nothing here is a new feature. It is the everyday loop, give context, ask precisely, refine, verify, applied end to end on a single deal so you can see the whole journey at once.

The task in front of you

A teaser came in last week and the CIM has just arrived. A partner wants an initial screen by Friday: is Atlas worth a closer look, and if so, what are the questions that decide it? You have the CIM, a financial model, a management deck, and an NDA. That is the raw material. Here is how you turn it into a judged view.

Step 1: Set up the deal Project

Resist the urge to drop the CIM into a fresh chat. A one-off question gets a one-off answer. A Project gives every conversation about Atlas the same standing context, so you never re-explain the deal.

Two things make this setup work. The instructions set Claude’s standing behavior for every chat in the Project: cite sources, separate fact from inference, flag gaps. The knowledge base holds the deal’s documents in one place, so any question you ask is answered against all of them at once. The full method is in Deal Setup; the thinking behind it is in The Two-Layer System.

Step 2: The first screen

With the Project set up, the first real question is the classic one: what is this business, and is it worth more of my time? You ask it inside the Project, so the CIM and the standing instructions are already in play.

The reply is a structured screen, grounded in the documents and honest about their gaps.

That last line is the tell of a good screen. Claude has not just summarized; it has noticed that a headline retention figure is quoted without a definition. That is the kind of gap that turns into a management question. A first screen of a 62-page CIM is exactly the moment to run at high thinking effort, so the reasoning has room to catch tensions like this one.

Step 3: Drill into the risks

A screen surfaces risks. The next move is to make each one specific enough to act on. You stay in the same chat, so nothing needs re-explaining, and push on the three the screen raised.

Take the three risks you flagged and make each one concrete.
For each: state exactly what the materials do and do not tell us, why it would matter to an investor, and the single sharpest question I should put to management to resolve it.
If the CIM contradicts the model or the management deck anywhere on these points, show me where.

This is the moment Claude earns its place. It caught a genuine tension, the founder-led narrative in the deck against the “scaled team” claim in the CIM, that is easy to miss reading the documents separately. The Due Diligence playbook turns this into a repeatable routine, and the Prompt Library keeps pressure-test prompts like the one above ready to reuse on the next deal.

Step 4: Check the numbers properly

The screen quoted growth and margin. Before any of those figures go near a memo, you want them computed, not estimated. You attach the model and ask for real calculation, which is where the analysis tool does the arithmetic for you.

Two useful outcomes. The growth and margin figures are now genuinely computed, so you can trust them in a deliverable. And Claude found a small reconciliation gap that becomes one more line on your management question list. The Financial Review playbook goes deeper on this.

Step 5: Frame the market

So far everything has come from inside the data room. To judge the investment angle, you need the outside-in view: how big is this market, who else plays in it, and what has moved lately. This is the one place you reach for web search. A search query leaves the Project, so it names the category, never Atlas or a figure from the CIM; Your Account & Confidential Data explains why.

Using current web sources, scan the market for B2B field-service software aimed at mid-market contractors.

Rough market size and recent direction of travel.
The main competitors, grouped as incumbents, challengers, and adjacent entrants.
Any notable funding, acquisition, or product news in this space in the last year.
Cite a source for each significant claim, and flag where the public picture is thin. Do not reference any specific target.

The scan gives you a public map of the landscape. Read alongside the CIM’s own market section, it lets you judge whether Atlas’s “fragmented market to consolidate” story holds up against what the outside world actually says. The full method is in Market Research.

Step 6: Draft the IC memo section

Now you have everything a first memo section needs: a sourced screen, concrete risks, computed financials, and an outside-in market view, all in one conversation. You ask Claude to pull it into a clean draft, produced as an artifact so you can refine it in place.

From here it is a conversation: “tighten the highlights for a partner,” “add a row on the management team,” “lead with the risks instead.” The document updates each turn, and you stay in control of the judgment. The Investment Memo playbook covers the full memo, section by section.

Step 7: Verify before it leaves your desk

You have a strong first draft in a fraction of the usual time. The final step is the one that never changes. Before this section carries your firm’s name, you check it.

1. Trace every figure to source

   Open each cited page and tab. Confirm the $60m, the growth rates, and the margins match the documents, not just the draft.

2. Confirm the flags are real

   Re-read the founder-versus-team passages and the churn wording yourself. A surfaced tension is a lead until you have seen it on the page.

3. Separate fact from inference

   Make sure the memo says what the materials state, marks what they imply, and names what they leave open. The reconciliation gap and the missing retention definition belong in the open items.

4. Apply your judgment

   Claude assembled the evidence. Whether Atlas is worth a closer look, and what to recommend, is yours to decide and yours to own.

The whole loop, on one deal

Step back and the pattern is the same one from your very first session, just applied with rigor across a real screen: a Project for context, precise questions, calculation for the numbers, a market scan for the outside-in view, a drafted section, and verification at the end.

That is Project Atlas, end to end. The features did not change between steps; the discipline carried them. Give Claude the right context, ask it precisely, let it do the synthesis and the math, and keep the judgment, and the final check, firmly your own.