Pillar · Due Diligence
Due Diligence Desk
Attest Dojo · Paul Sully · AIpreneur
Osaka, Japan · worldwide by email · tax included
Scoped, source-backed research for go/kill decisions on a vendor, agent, or deal. The result metric is hours and days saved: a scoped, source-graded handoff instead of open-ended research burn. No fake client logos. No invented EBIT %.
- Hours saved deciding go / kill
- Source-graded findings with citations
- What was checked and what was not, in writing
Time saved
Stop the research tax.
Keep a record.
Endless tabs without a handoff is a time tax. Due Diligence stops the burn and returns a scoped, source-graded memo you can act on: go, kill, or ask the next hard question. Hours recovered is the result. The handoff is the record.
Simple demo · your numbers, not a client claim
Hours toward a decision with a record: 8 saved vs open-ended burn (your inputs).
Illustrative math only. Real scope and price are agreed by email before work starts.
Time tax
Honest waste calculator
(optional).
Quiet agent failures also burn calendar. Rough yen cost of that tax, compared to the Sprint price of ¥600,000. Not a promise that Sprint “pays for itself” in X weeks. A framing tool.
Minutes × failures/week × yen/hour
About ¥288,000 / month of time tax at your inputs. Sprint price: ¥600,000 (tax included).
The work
Scoped research with a record.
Due Diligence research is scoped work for decisions that need a record. We grade sources, separate claim from inference, and say what is still unknown. Intake is email only. Scope, price, and what we will and will not do are agreed before work starts.
- Scope · what we will and will not cover, and which records we can reach
- Work · registry and source checks with citations, sanctions in the subject's own language
- Handoff · a written pack that states what was checked and what was not
Lanes
What the research covers.
- Screen · identity lock on the subject, sanctions and watchlist checks, jurisdiction coverage, and baseline findings, each with its source
- Standard · a fuller entity pack: registry records, officers and structure, adverse media, and a scored summary with sources
- Enhanced · deeper multi-source research on the same subject, claim by claim, with a coverage statement
- Ownership and control · ownership hops on company filings and public registers, each hop cited, each finding marked found, partial, or a gap; a closed register is reported as a gap
- Monitor · periodic checks on the same subject, and a written note when something material changes; not every change will be caught
- Embed · scoped support across these lanes for a defined period
- Crypto · referral to a specialist on-chain partner; not in-house work
A check reports what the sources show. It is not a clearance of any person or company.
The evidence
What you can check in a pack.
Eight things about our packs that you can verify yourself. Each one describes the record we leave, not a claim about our methods.
Evidence 01
Every point carries one of three labels.
Each point in a pack is marked verified against a source, claimed (a party says so, and the source is named), or unverified (with the reason: nothing qualifying found, source unreachable at capture, sources disagree, or a gap in coverage). Nothing is rounded up. Before delivery, a pack goes through a verification gate and a separate review of the finished pack.
You see it in: the label beside each point, and the verification result on the pack cover.
Evidence 02
Findings are checked against their source before they harden.
Claims are matched back to the document they came from and re-checked before they are labeled verified against a source. Anything that cannot be re-checked stays claimed or unverified.
You see it in: the citation attached to each finding.
Evidence 03
Every captured document is fingerprinted.
Each document is hashed the moment it is captured, with the capture date and where it came from. A pack can show what was seen, when, and from where. This is a record of our capture, not a forensic or legal custody claim.
You see it in: the stamp beside each point verified against a source: source type, capture date, and fingerprint.
Evidence 04
Your matter stays under our control.
Case files, evidence, and pack contents are not sent to public AI services and are not published anywhere. What we gather for you stays with the engagement.
You see it in: the handling and privacy note in the engagement letter and in the pack.
Evidence 05
Names are matched and entities are read in the subject's own language and script, starting with Japanese, Spanish, and Filipino forms.
The sanctions check was tested against Japanese-name variants (kanji and kana) and against Filipino name forms: the ng̃ spelling, ñ, the y/i conjunction, name particles, reduplication, and local company suffixes. Entity recognition was tested on official Spanish pages and reads person and organization names in the subject's own language.
You see it in: the sanctions and entity sections, which cover Japanese, Spanish, and Filipino name forms instead of skipping them.
Evidence 06
Every pack states what was checked and what was not.
Coverage is reported honestly: what was found, what was not searched, and where paid or official portal access is still needed to go further.
You see it in: the "what we checked" section, which reads as gaps you can act on, not as prose.
Evidence 07
Deep work tracks every claim in a ledger.
In deeper engagements, each claim carries its source, its status, and the date it was checked against that source. Nothing is quietly promoted from claimed to verified against a source.
You see it in: the claim ledger and source grading in deeper engagement packs.
Evidence 08
Thoroughness is measured, not claimed.
Screens are scored against known-answer benchmarks, with precision and coverage recorded in a results log. The word "thorough" is earned by a number, not by an adjective.
You see it in: the documented quality discipline behind every screen.
What we do not claim
Clear boundaries.
We do factual research over public records and published sources. We are not lawyers, and we give no legal advice or legal conclusion. We do not conduct interviews, surveillance, or pretext calls, and we do not contact the subject or anyone around them. We do not make, or supply material for, eligibility decisions on employment, credit, tenancy, or insurance. We claim no licence, certification, or legal clearance in any jurisdiction. Quality of earnings, audit-style assurance, and regulatory compliance opinions stay with the specialists who own those lanes. Where a matter needs commercial databases, court dockets, or local counsel, we add them for that matter, and the pack says so.
Findings support a decision. They are not legal advice and not an audit opinion. Matters that would need a private investigator licence or registration (in Japan, a detective business notification) are not accepted and are referred to a local provider.
How it starts
Three steps. No sales funnel.
- Email staff.duediligence@attestdojo.com with the entity or matter and the decision you are trying to make, and the desk replies with scope and price
- We agree scope, price, and what we will and will not do, by email (engagement letter + conflict check)
- If it fits, work starts. You receive a written pack that states what was checked and what was not
Next
Scope and price by email first.
Due Diligence fees are agreed by email for the scoped matter. No public fixed price band on this desk. A person replies. There is no self-serve checkout on this page.
About
Paul Sully · AIpreneur · Attest Dojo 代表 / エージェント信頼性コンサルタント
SI work with a record. Agents that stay consistent. Research with sources. Osaka. Systems that run, not demos that impress.