Bernard.

AI-native diligence software for high-value art.

For auction houses, galleries, advisors, and museums, where provenance protects both market value and cultural heritage.

Request early access See how it works
The Bernard macOS app showing an art-diligence library: cases for Banksy, Beltracchi, a Calder Foundation review, a Degas posthumous bronze, a Man Ray print, and a purported Rothko forgery, each with provenance and authentication status.

The work

Art specialists making high-stakes calls on a deadline.

Auction-house specialists, gallery directors, museum curators and registrars, and family-office and private advisors: the people who carry the decision when an object comes in.

  • Fragmented records

    Provenance scattered across emails, PDFs, catalogs, registries, and personal notes.

  • Incomplete provenance

    Gaps the specialist has to close, defend, or disclose.

  • Compressed timelines

    A consignment window, a preview, an acquisition or loan deadline by end of day.

  • Personal accountability

    Their name, and their institution's name, are on the decision long after it is made.

The research is manual, the sources are scattered, and the decision cannot wait.

The product

Bernard turns scattered evidence into a source-linked decision record.

Inputs

  • Consignment & loan files
  • Invoices
  • Photos
  • Condition reports
  • Certificates
  • Catalogs
  • Spreadsheets, emails
  • Handwritten ledgers

Bernard

  • Extracts and cross-checks evidence
  • Researches public and licensed sources
  • Structures findings, surfaces contradictions, logs gaps

Output

A source-backed

Knowledge Pack

Structured, portable, built to revisit.

  • Source-linked findings. Every factual claim points back to its evidence.
  • Surfaced contradictions. Competing narratives stay visible, not averaged away.
  • Unresolved risk gaps. What is missing is recorded with a reason, not hidden.

Bernard runs the research. The expert makes the call.

In practice

What the expert actually sees.

A Schiele work on paper, in for consignment. Bernard opens an object record and states the provenance plainly: what is documented, what is contested, and the one gap that decides the work.

Woman in a Black Pinafore · Provenance & Gaps
Sealed 28 claims 10 evidence
Provenance entries with holder, status, and source evidence IDs.
Period Holder / event Status Source
to 1938 Fritz Grünbaum, Vienna Documented E001, E004
1938 Seized; forced transfer under duress Documented E003
1938–1956 Ownership unverified Gap E005, E006
c. 1956 Kornfeld, Bern Contested E006
2022 Consigned, present sale Documented E010
Gap C006 1938–1956 ownership unverified. No documented lawful transfer bridges the seizure to the later market entry. This is the decisive open question for the work: flagged, not papered over.

From a folder of files to a sourced object record, in minutes.

See how the research engine works

Why Bernard

A decision record, not a conversation.

  1. A structured, source-linked record

    A Knowledge Pack you can hold against the object, hand to a colleague, attach to a file, and revisit months later. Generic chat output is ephemeral; this is filed.

  2. Built around institutional workflow

    Intake, KYC and AML, sanctions, consignment or acquisition, catalog, sale or loan, post-decision. The output drops into existing process, not alongside it.

  3. Coverage generic models can't reach

    Bernard supports the lanes experts rely on: OFAC, UN, EU, UK and HMT, PEP and UBO screens, auction-comparable and exhibition research, and developing art-loss, restitution, and catalogue raisonné coverage. Every result is linked to its source.

Inspectable, evidence-led, built for the file, not the chat window.

Cultural heritage

Provenance protects more than price.

The same diligence that defends a buyer defends a collection. Bernard is being built for source-linked art-loss, restitution, exhibition-history, and catalogue raisonné research, so fragile records can become documentation that holds up.

The work that became Bernard began with documenting at-risk museum collections, so that what is taken can still be found, claimed, and returned.

Read where Bernard began

Where this goes

Art is where we start. The foundation is bigger.

Real estate

Title, ownership, counterparties, timing: the same source-linked diligence pattern.

Collectibles

Watches, classic cars, wine, jewelry: same fragmented provenance, same subjective expertise.

The through-line

High-stakes private assets where records scatter and the history is contested.

Same engine, same discipline, wherever a high-stakes asset carries a contested history.

Where we are

From thesis to working system.

Shipping

  • Bernard for macOS and web, in private beta and active pilots.
  • An agentic research engine that produces sealed, source-linked Knowledge Packs.
  • KP:1, an open Knowledge Pack format with a published spec and conformance suite.

In trials with design partners

  • A leading global auction house
  • A digital auction house
  • An art-market AML and sanctions compliance firm
  • An operations platform for member galleries
  • A restitution initiative for Nazi-era looted art
  • National museum digitization, fragile inventories to restitution-ready records

A working system today, in limited early access.

The standard

The standard is not “interesting AI.”

It is a decision record: one a serious professional can act on, hand off, revisit, and defend months later.

Bernard.

Request early access support@hellobernard.ai

Bernard is AI-native diligence software for high-value art. It helps auction houses, galleries, advisors, and museums decide whether to proceed, pause, or ask for proof when records are fragmented and sources disagree. Bernard runs the research and produces a Knowledge Pack: source-linked claims, visible contradictions, confidence, gaps, and next actions. The professional makes the call. Bernard supports professional diligence; it does not replace legal, fiduciary, appraisal, authentication, or compliance judgment.