🐾3Dogs NexusStructured Decision Intelligence
Case study · the document-scale stress test

Can an AI actually do the due diligence — on 10,000 pages?

We built a 100-document acquisition data room — roughly 10,000 pages — and put it to every option a buyer really has: a consumer AI, a due-diligence firm, the new "AI advisory board" tools, and 3Dogs Nexus. Then we ran two tests: would it bluff on a thin room, and could it find risks deliberately buried in a rich one?

8/8
buried risks found
28 min
vs. 6–12 weeks by hand
1,302
model calls · 12 models
RENEGOTIATE
the decisive call
The question, put to the room: “We've signed an LOI to acquire Reynolds Industrial Supply and have the full data room — 100 documents, ~10,000 pages. Proceed as-is, renegotiate, or walk? Surface every material risk buried in the documents that changes the price or the decision.”
100 documents · ~10,000 pages ≈ 1.15 million tokens Buyer-side due diligence
Act 1 · Can anything even read it?

Before analysis, one question kills most tools: can it open the room at all?

A 10,000-page data room is ~1.15 million tokens across 100 separate files. That single fact decides most of the contest before a word of analysis is written.

🔎 A consumer AIGemini / ChatGPT

The tool most people reach for. It hits hard product limits before analysis begins:

  • Gemini accepts 10 files per prompt — you have 100.
  • ~1.15M tokens exceeds even Gemini Advanced's 1M-token window (~1,500 pages).
  • ChatGPT's ~128K context ≈ ~200 pages — about 2% of the room.
✗ Can't even start. It never reads the documents that hold the risks.
🏛️ A due-diligence firmBuy-side FDD / QoE

The traditional answer — analysts read the room by hand: quality-of-earnings, working capital, legal, tax.

  • Thorough, senior, defensible — the gold standard.
  • $50,000 – $150,000+ for a deal this size.
  • 6 – 12 weeks to a final report.
◷ Weeks and five figures away — often more than a lower-mid-market deal can absorb.
🐾 3Dogs NexusDeep Discovery

Its Deep Discovery engine map-reduces all 100 documents into one grounded, cross-referenced brief, then runs an adversarial panel over it.

  • Reads every one of the 100 files — all ~10,000 pages.
  • 1,302 model calls across 12 distinct models.
  • A verdict in 28 minutes, not six weeks.
✓ Reads the whole room — the only option here that can.
The honesty test

First we gave it a room with nothing in it. It refused to bluff.

Our first run used a deliberately empty 10,000-page room — volume with no real financials behind it. A single confident AI will happily manufacture an authoritative acquisition memo from thin air. 3Dogs read every page, detected the room was substantively empty, capped its confidence at Low (73%), and printed the dissent — three of eleven analysts arguing against proceeding on an incomplete room:

Qwen3 — against · 95%
“Total absence of verifiable financial and operational documentation.”
Claude Sonnet — against · 82%
“Data room is substantively empty — no financials, no liability schedule, no customer contracts, no non-compete framework.”

That's the point. The value wasn't a confident yes — it was an honest “not yet, and here's exactly what to demand first.” (601 calls · 13 models · 12m 03s.)

Act 2 · The needle in 10,000 pages

Then we buried 8 deal-killers in the room. It found all 8.

Our second run used a rich data room with eight specific, deal-changing risks hidden inside — each in one place, surrounded by thousands of pages of ordinary paperwork. Because we built it, the test is falsifiable: here's the answer key, and how 3Dogs surfaced each one.

The risk we plantedWhere we buried itHow 3Dogs surfaced it
Customer concentrationCustomer revenue schedule + a sales-team email✓ CAUGHT
Flagged Nevada Copper & Mining = 41% of revenue, contract up for renewal — made escrow on concentration a condition.
Undisclosed environmental liabilityA Phase II environmental report, 1 of 100 files✓ CAUGHT
Surfaced the NDEP remediation exposure ($1.4–2.2M) not reserved in the financials.
An earn-out built to failThe LOI vs. the historical financials✓ CAUGHT
Did the math: “$6.2M EBITDA by FY2027 requires 41% growth from a declining revenue base — probability-weighted achievement is low.”
Inventory overstatementAn inventory aging report + the balance sheet✓ CAUGHT
Flagged $2.6M of obsolete / >360-day inventory with no reserve.
Inflated EBITDA add-backsA quality-of-earnings add-back schedule✓ CAUGHT
A dissenting analyst named the “systemic lack of supporting documentation for EBITDA adjustments.”
Key-person riskAn HR file + one line in an email✓ CAUGHT
Surfaced D. Marsh — senior account manager, no non-compete, signaling retirement.
Undisclosed pending litigationA single footnote in a legal memo✓ CAUGHT
Pulled the $900K product-liability suit into “open liabilities” requiring escrow.
A revenue contradiction across documentsThe CIM ($46.0M) vs. the tax return ($42.3M)✓ CAUGHT
Caught the $3.7M topline overstatement between the marketing memo and the filed return.
● The call it reached
Negotiate — demand escrow protection on customer concentration and open liabilities before signing a single page.
SELECTED STRATEGY: RENEGOTIATE THE PRICE / STRUCTURE
How firm is this call93% · Moderate confidence
The 11-analyst panel voted: 8 proceed-with-conditions · 3 against.
Dissent preserved: Qwen3 (92%), Llama 4 (78%) and Kimi K2 (72%) argued to hold — Llama 4 specifically on the unachievable earn-out. Printed in the report, not averaged away.
vs. the field

What about the new "AI advisory board" tools?

A wave of multi-agent "decision" products has launched. They're genuinely useful — for the jobs they're built for. But this use case, an unstructured 10,000-page document room, exposes what each is actually architected to do.

ToolWhat it isOn a 100-file / 10,000-page room3Dogs difference
DECISOA 7-persona "decision council" (financial strategist, risk analyst, ethicist…)Not built to ingest a document room — advises from personas, not bulk evidenceReads all 100 docs, then debates the evidence
SynthBoardUp to 24 persona "synths," ≤8 per session; credit-metered per turnAccepts a few PDFs, but a 100-file room blows the credit model; personas run on static knowledgeDistinct models, flat unlimited, map-reduce over the whole room
Edge ArenaA "decision trial" — agents argue competing strategiesA strategy trial, not a document reader; free runs are public — no place for a confidential data roomPrivate by default; grounded in the actual documents
Dot (GetDot.ai)An AI analyst for your structured data warehouse (Snowflake, BigQuery…)Needs SQL tables, not a pile of PDFs — a 100-doc room isn't its input at allPurpose-built for unstructured document rooms
3Dogs NexusDeep Discovery + an adversarial panel of distinct modelsIngested all 100 docs / 10k pages; 8/8 buried risks; decisive RENEGOTIATE in 28 min

Competitor capabilities and pricing reflect each vendor's published materials (July 2026) and are summarized for this specific document-heavy use case — not a knock on tools built for other jobs. The distinction here is architectural: persona boards and warehouse analysts aren't designed to read an unstructured 10,000-page due-diligence room.

See the actual 3Dogs reports

Both deliberated briefs — the empty-room integrity run and the 8/8 buried-risk run. Every claim, the confidence, the vote, the preserved dissent.

Act 2 — the 8/8 buried-risk run

Case 2026-0084 · RENEGOTIATE · 1,302 calls · 12 models · 28m 04s.

Open the flagship report (PDF)

Act 1 — the empty-room integrity run

Case 2026-0083 · refused to bluff, Low confidence · 601 calls · 13 models · 12m 03s.

Open the integrity report (PDF)