Early red flags within 48 hours of an evidence-complete kickoff; standard final pack in five business days.
AI-native technology due-diligence service
The diligence,
delivered.
Give us the evidence room and deal mandate. DD Memo delivers the reviewed, evidence-cited verdict, quantified deal implications, and IC-ready decision pack - in days, not weeks.
You buy an accountable outcome - not seats, tokens, or another tool for your team to operate.
6specialist AI workstreams
7required challenge stages
48hearly-red-flag target
5 daysstandard final-pack target
§ 01 — THE PRODUCT
Watch the diligence get done.
Not a copilot waiting for prompts. A production system that reads the room, attacks its own first answer, and lets code - not narrative - decide what can be claimed.
Drop the room in. It files itself.
Upload the target's material - folders, zips, scans. DD Memo reads each document and routes it to the right workstream by content, because every specialist reads only its own domain. Low-confidence routings are queued for a one-look human confirm.
Per-deal isolation · information barriers · OCR for scans · whole-room zip intake
SOC2_TypeII_bridge_letter.pdfrouted by content · confidence high→ SECURITY
architecture-overview-2026.pdfrouted by content · confidence high→ ARCHITECTURE
vendor-agreements.zipexpanded · 14 files routed individually14 ROUTED
scan_047.pdfOCR · low routing confidenceREVIEW
agents read only their own workstreamevery read is audit-logged
Then separate challengers attack the answer.
Six specialists produce the first view. Independent functions then try to refute material findings from their cited sources, rebuild the key numbers from first principles, and test every claim against your deal mandate. What survived, what changed, and what remains unknown all travel with the pack.
Evidence refutation · quantitative re-derivation · mandate audit · injection canaries
“No unremediated criticals”pentest report contradicts · 2 open criticals foundREFUTED
ARR $12.4Mrebuilt bottom-up from billing exports · Δ −2.1% vs managementUPHELD
“DR tested annually”no test evidence in the roomINSUFFICIENT
“SOC 2 Type II complete”bridge letter only · report in progressDOWNGRADED
challenge verdicts ship with the memonothing is smoothed over
And code decides what can be claimed.
Every value in the memo wears a provenance label - Verified, Derived, Inferred, Illustrative, or Unknown. A citation that does not resolve to a real passage is downgraded automatically. An open critical or a failing mandate gate clamps the recommendation, reason shown. A named engagement lead signs before anything ships.
Deterministic gates · provenance on every value · hash-chained audit trail
Remediation cost, year one: $1.8Mcited to pentest + vendor quotesVERIFIED
Integration run-rate delta: $640kformula shown · inputs citedDERIVED
Vendor exit exposureanalyst judgement · reasoning shownINFERRED
GO GATE · ENFORCED IN CODE
2 open criticals → recommendation clamped to CONDITIONAL-GO. The model cannot narrate past a red flag.
2 open criticals → recommendation clamped to CONDITIONAL-GO. The model cannot narrate past a red flag.
signed by a named engagement leadsha-256 hash-chained record
§ 02 — WHAT CHANGES
Focus on the decision,
Focus on the decision,
not the document pile.
Every material claim cited, every value provenance-labelled, every challenge verdict on the record.
Certifications, thresholds, dealbreakers - code gates, not suggestions in a prompt.
Managed isolated, your VPC, or fully local and air-gapped on a local-LLM backend.
§ 03 — Where we stand
We are opinionated about how AI belongs in diligence.
POV / 01
Never trust one model answer.
A single pass from a frontier model is a plausible draft, not diligence. Scrutiny has to be part of production: independent refutation, re-derived numbers, a mandate audit, and deterministic completeness checks - before the buyer ever sees a conclusion.
POV / 02
Accuracy is an invariant, not an aspiration.
Estimates must never dress up as facts. Every value carries how-it-is-known; unresolved citations downgrade automatically; illustrative numbers are excluded from the underwriting math. Blank never means fine - a gap is a finding.
POV / 03
The software is the labor. The service is the guarantee.
Nobody wants more software to operate. You buy the completed diligence outcome; the production system is how it arrives fast and consistent, and a named engagement lead is accountable for what ships.
§ 04 — The lowest-risk first purchase
Start with a deal you already closed.
We freeze DD Memo's analysis before seeing the prior diligence answer, then compare risks found, misses, evidence limits, cost estimates, and decision usefulness side by side.
Paid retrospective proof
- Agree the original evidence cutoff and success measures.
- Run the full production and challenge method.
- Adjudicate against prior diligence and available post-close outcomes.
- Credit 100% of the proof fee to the first live engagement signed within 60 days.
§ 05 — ECONOMICS
Custom fixed fee. Materially lower than labor-heavy diligence.
There is no public rate card. Each proposal documents the comparable scope, evidence assumptions, delivery configuration, fixed fee, and calculated client saving.
§ 06 — SECURITY
Run where your security team requires.
Managed isolated
A dedicated DD Memo-run environment with agreed no-training, retention, and access controls.
Customer-controlled
DD Memo operates inside your cloud account under your access and egress policies.
Fully local / air-gapped
The engagement runs on a local-LLM backend (vLLM / Ollama) inside your boundary - no external model egress at all - for strict data-residency and no-egress mandates.
Every configuration uses per-deal isolation, role enforcement, tamper-evident audit, watermarked exports, evidence citations, and no model training on client evidence. The AI backend is a per-client choice: the Anthropic API we operate under no-training terms (zero-data-retention available) or a local LLM inside your boundary.
§ 07 — QUESTIONS BUYERS ASK
Asked, answered, on the record.
How is this different from AI document search like Hebbia or Toltiq?
Those tools make documents queryable - a human still has to ask the questions, assemble the answers, and defend them. DD Memo delivers the completed diligence: the structured questionnaire answered, findings challenged, the mandate enforced in code, and an IC-ready decision pack signed by a named lead.
Can our security team approve this for live-deal MNPI?
That is what the delivery configurations exist for. Per-deal isolation, information barriers, a hash-chained audit log, watermarked exports, and no model training on your evidence apply everywhere; the engagement can run in our isolated environment, inside your VPC, or fully air-gapped on a local model. A security whitepaper and AI governance model card are available on request.
Does the AI make the investment decision?
No. The system produces decision support: the model proposes, deterministic code enforces the evidence rules and your deal criteria, and a named engagement lead reviews and signs. Your investment committee keeps the final call - and the pack shows its work so the IC can challenge any line of it.
What if the evidence room is messy or incomplete?
Messy is normal. Documents are auto-routed by content, scans are OCR'd, and whole-room zips are expanded file by file. What is missing becomes explicit: coverage gaps and open evidence requests are reported as findings, never silently skipped - so the memo tells you what it could not know.
What does an engagement cost?
A custom fixed fee for the completed outcome - no seats, tokens, or usage meters. When a comparable traditional proposal or invoice exists, the quote documents the baseline and the calculated saving. The retrospective proof is a privately quoted paid engagement, credited in full against the first live engagement signed within 60 days.
How fast is an engagement?
Early red flags are targeted within 48 hours of an evidence-complete kickoff, the standard final decision pack within five business days, and complex carve-outs in seven to ten. The pace comes from the software doing the production - not from skipping the challenge stages.
Make the first test falsifiable
Bring one closed deal.
See what DD Memo catches.
No software rollout. No free-trial theater. A paid, scoped proof with a decision pack your team can challenge.