Product

Diligence with a defensible record.

Bernard helps art specialists, gallery directors, museum curators and registrars, and advisors decide whether to proceed, pause, or ask for proof, when provenance is thin, sources disagree, and the window is short.

The Bernard macOS app: a library of art-diligence cases including Banksy editions, a Beltracchi forgery, a Calder Foundation review, a Degas posthumous bronze, a Man Ray print, and a purported Rothko forgery review.

Decision pressure

Concrete questions. Short windows.

Bernard is built for the specialist who carries the decision. The institution benefits.

  • Postwar work on paper

    Approve before the hold expires?

    Provenance is thin, condition is pending, and restitution and export checks have to land before Friday.

    Consignment · Friday

  • Museum acquisition

    Accept the gift, or ask for more?

    An offered collection needs provenance, title, and art-loss screening before the board meets.

    Acquisition · 2 weeks

  • Cross-border loan

    Clear the object for exhibition?

    Ownership history, sanctions exposure, and loan conditions need one defensible view before the catalogue locks.

    Loan · 10 days

Research loop

Each evidence pass replans the next.

Bernard runs diligence as a working record. Each pass checks claims, adds proof, preserves conflicts, and keeps gaps visible until the next move is clear.

The working record

Claims, evidence, source lineage, conflicts, gaps, confidence, views, and history stay in one Knowledge Pack. Every pass reads it and writes back to it.

  1. 01

    Read the record

    Subject, documents, current claims, evidence, gaps, and the decision window.

  2. 02

    Find gaps

    Missing proof and weak claims are ranked by decision risk.

  3. 03

    Choose checks

    Targeted questions and primary-source paths are selected.

  4. 04

    Test evidence

    Sources are checked against the open claims and known conflicts.

  5. 05

    Update confidence

    New proof, contradictions, confidence changes, and citations enter the pack.

  6. 06

    Name the next move

    Proceed, pause, ask for proof, or keep an open gap visible.

Knowledge Pack

Confidence in the moment. Evidence for future inquiry.

A Knowledge Pack is the decision record Bernard produces. It preserves what was claimed, what evidence supports it, where sources disagree, what remains unproven, how confidence changed, and what action is justified next.

A Bernard Knowledge Pack open to Provenance and Gaps for a Schiele work: a source-linked table of holders and events, each with a documented, contested, or gap status, and a plainly stated 1938 to 1956 ownership gap.
Provenance and Gaps for a Schiele consignment. Every entry is source-linked; the gap is named, not hidden.

Example next action

Claim C014     Lender released the work
Status         Unproven, release letter missing
Confidence     Low
Evidence       E021 (consignment file), E022 (lender schedule)
Next move      Ask for proof before approval

The reviewer sees the recommendation and the reason for it. The next move can be challenged, updated, or cleared when proof arrives.

Anatomy

What a Knowledge Pack is made of.

A pack is the research record itself: claims, evidence, confidence, contradictions, views, and integrity metadata in one portable folder or sealed archive.

PACK.yaml
Identity, domain, author, sensitivity, dependencies, and views.
claims.md
Atomic findings with confidence, type, evidence IDs, date, depth, and relationships.
evidence.md
The source ledger, where every cited ID resolves to its lineage.
views/
Readable outputs for people, derived from the authoritative claims and evidence.
attachments/
Optional source material and supporting documents.
history.md
Superseded claims, revisions, and reasoning changes.
.kpack
The sealed archive format for transfer, backup, verification, and version lineage.

AI systems

Load structured claims, evidence, entities, contradictions, and history as reasoning context.

Professionals

Read views, risk summaries, source notes, and open actions while preserving the underlying record.

Organizations

Keep a reusable, inspectable record across handoffs, revisions, and later questions.

Integrity

Hashes show whether content changed. Sealing is the integrity layer future certification can build on.

Coverage

Your private data, plus the public record.

Most diligence tools only see public data. Bernard connects the private material you bring with public-record research and developing specialist coverage, every result linked to its source.

You bring · private

  • Capture browser for any site you are authorized to use
  • Research portals and subscription databases you license
  • Local files: PDFs, invoices, condition reports, certificates
  • Internal records: past deals, correspondence, your own notes

Bernard researches · public

  • Sanctions / PEP: OpenSanctions (OFAC, UN, EU, UK/HMT)
  • Ownership & corporate: OpenCorporates, Companies House
  • Courts & filings: CourtListener, SEC EDGAR
  • Property: HM Land Registry
  • Academic & researcher: OpenAlex, arXiv, ORCID
  • Public web: auction results, exhibition history, press

Bernard is building · specialist

  • A proprietary art-market dataset, independently built
  • Art-loss register: looted and stolen-work screening
  • Catalogue raisonnĂ©: authoritative work catalogs
  • Auction-house data: results and provenance

Private, public, and proprietary, synthesized into one source-linked record.

Discipline

Breadth is not proof. Lineage is.

Bernard separates retrieval breadth from evidentiary strength. Repeated mentions of the same source do not raise confidence; independent source lineage can.

  • Shared origin counts once

    Three tools finding one article is still one source. Repeated mentions of one underlying origin are treated as one source.

  • Primary records anchor claims

    Filings, registries, catalogues raisonnés, and source documents anchor claims wherever they are available.

  • Gaps stay visible

    Missing primary routes remain visible as gaps, open actions, or contradictions, never smoothed over.

Local-first by design: the Knowledge Pack is the durable record you keep. Each pass works from a scoped slice and returns cited findings, not opaque summaries.

Markets

Art is the proving ground. The foundation is bigger.

Bernard starts where mistakes are public, evidence is fragmented, and the tooling is decades behind. The research lens changes by asset class; the record stays one thing.

Art-market diligence

Provenance, condition, ownership, market context.

A source-backed provenance view, condition gaps, a comparable range, and the questions to settle before the hold expires.

Real-estate assets

Title, obligations, planning, comps, counterparty risk.

An offer-readiness view, red and amber flags, comparable support, and the next requests before the window closes.

Financing and counterparties

Authority, pledge status, identity, release proof.

Proceed, pause, or ask for proof, with every material claim tied to a source.

Plain answers

What Bernard is for.

Can my team inspect the record without trusting a black-box summary?

Yes. KP:1 keeps the record in plain-text files, so a colleague, reviewer, or system can inspect claims, evidence, contradictions, confidence, gaps, and next actions directly.

Will my files train public AI models?

No. Bernard is built around local-first Knowledge Packs and scoped research tasks. Evidence you upload is not used to train public AI models.

How is Bernard different from a research agent?

A research agent may answer a question. Bernard preserves the decision record: what was claimed, what evidence supports it, where sources disagree, what remains unproven, and what action is justified next.

What does a Knowledge Pack actually contain?

Decision context, claims, evidence, contradictions, confidence, gaps, next actions, readable views, and revision history, in a portable .kpack folder.

Who makes the final call?

The responsible professional. Bernard preserves evidence and open questions; it does not make or replace the decision.

See it work

From a folder of files to a sourced record.

Read the transcript

Introducing Bernard.

AI-native diligence for professionals who handle sensitive, high-stakes work.

When a good opportunity comes in, everyone wants to move. The price looks fair. The documents look complete. The buyer is ready.

This is when teams face a bad choice. Move fast and hope nothing is missing. Or slow the deal down while someone checks.

Bernard gives you a third option. It does the careful check right there, in minutes, while the opportunity is still live.

It reads the details. It checks the names. It flags what does not add up.

Here is Larkspur Study. It looks ready to consign.

But Bernard finds three problems. The signer is not the trust. The artwork is already listed on a lender schedule. And the release letter is missing.

So the next step is clear. Do not approve it yet. Ask for proof the signer can act for the trust, and proof the lender has released the artwork.

The same careful check works across the whole library.

A townhouse looks ready to exchange, until the title notice is still on the register.

A new person or company looks ready to send funds, until the passport name does not match.

A loan against artwork looks ready to fund, until an old pledge is still open.

Bernard is not there to stop the work. It shows what has to be fixed, so the work can keep moving with proof.

And sometimes, after the same checks, everything lines up.

The authority is clear. The release is in. The payment route is approved.

That is the green light. Now the answer can be yes.

Bernard does not make the final call for you. It keeps every claim tied to a source, so you can see the proof yourself.

Bernard runs locally. Your files stay on your device. Your work stays private.

You stay in charge. Bernard finds the facts, connects the sources, and gets out of the way.

Bernard. See what matters before you move.

Bernard runs the research. The expert makes the call.