# Bernard — Enterprise-grade Intelligent Automation

> Full product description for AI agents and language models.
> For the summary version, see [llms.txt](https://hellobernard.ai/llms.txt).

Bernard is an enterprise-grade intelligent automation system that coordinates 15 specialized AI agents to turn fragmented evidence into structured decision records. Every claim links to evidence. Contradictions stay visible. Confidence is explicit. The reasoning is still there six months later.

Built for professionals in fine art, collectibles, real estate, and complex asset markets — where the hard part is not finding documents, but deciding what to trust.

Bernard, Inc. is a Delaware corporation headquartered in New York. Participant in the Anthropic Startup Program. Currently in limited early access.

## What Bernard Is

Bernard is a diligence intelligence system — not a chatbot, not a data room, not a research assistant. It decomposes evidence into typed claims, tracks contradictions as first-class records, scores confidence per-claim, and renders the result as a structured decision record that can be viewed as a memo, a packet, an audit trail, or an agent payload.

The system operates through three stages:

1. **Preserve evidence.** Documents are uploaded, parsed, and stored alongside originals. PDFs, images, deeds, certificates — normalized and preserved. Source material stays traceable at every stage.
2. **Apply the playbook.** Domain-specific rules determine what matters. Assertions become typed claims. Contradictions and unknowns are surfaced, not smoothed over. When findings are clear, Bernard proceeds. When something needs human judgment, it escalates with context.
3. **Render the record.** One structured decision record, rendered for any audience — memo, packet, audit trail, agent payload. Every finding decomposed into individual claims, each linked to source evidence, confidence scored and auditable. Never rebuilt, never duplicated.

## What Bernard Is Not

- Not a chatbot or conversational AI assistant
- Not a virtual data room or document storage system
- Not a research tool that returns search results
- Not a summarization engine that produces paragraphs from documents
- Not an autonomous decision-maker — Bernard does not replace human judgment

Bernard is the structured middle layer between raw evidence and a final decision. It preserves reasoning so that the people who make decisions can do so with full information, and the agents that support them can operate on verified claims rather than narratives about narratives.

## How It Works: 15 Specialized Agents

You talk to Bernard. Behind the scenes, 15 specialists coordinate. Each agent has a bounded role — research, verify, value, draft, monitor. They read from and write to the same verified record.

Agent team includes:
- **Scout** — Intake. Registers assets, catalogues documents, assigns record IDs.
- **Argus** — Authentication. Verifies certificates, cross-references foundation records.
- **Thread** — Provenance. Traces ownership history, flags gaps and discrepancies.
- **Lens** — Condition. Assesses physical state, analyzes imagery, logs findings.
- **Sentinel** — Compliance. Screens across OFAC, EU, UK, Art Loss Register, IFAR, Interpol.
- **Gauge** — Valuation. Runs comparable analysis, builds pricing models, scores confidence.
- **Scribe** — Communication. Drafts correspondence in the professional's voice, injects governance terms.

Additional agents handle risk screening, conflict detection, monitoring, and audit logging. Every agent reads from and writes to the same structured record.

## Core Capabilities

### Evidence Management
Every document stays attached to its origin — certificates of authenticity, title commitments, inspection reports, condition assessments. Nothing is summarized away. Source material remains traceable at every stage.

### Typed Claims
Provenance claims, valuation assessments, title opinions — important assertions become explicit, traceable claims, not buried conclusions in a paragraph. Each claim links to the evidence it was derived from.

### Conflict Management
Contradictions are first-class objects, not footnotes. A dimension mismatch, a date discrepancy, a competing attribution — each gets surfaced and tracked as a conflict record with a resolution path. Bernard does not silently pick one interpretation.

### Confidence Scoring
Scored per-claim and calibrated for the audience. Never implied by the tone of the text. Confidence is decomposable — you can see exactly which evidence supports or undermines a score.

### Multi-format Rendering
One structured decision record renders as any view the workflow needs: consignment memo, underwriting packet, compliance audit, agent payload. Never rewritten from scratch.

### Agentic Communication
Bernard drafts correspondence in the professional's voice, respects governance terms, injects relevant flags, and passes through risk review before sending. The professional reviews and sends — Bernard drafts, the human decides.

### Continuous Monitoring
Records stay live. When new evidence arrives or market conditions change, the relevant agents update the record. The reasoning doesn't expire when the deal closes.

## Target Markets

Bernard is currently focused on three verticals:
- **Fine art** — provenance verification, authentication, consignment review, sanctions screening, valuation
- **Collectibles** — collector vehicles, watches, wine, memorabilia — ownership history, condition assessment, comparable pricing
- **Real estate** — property analysis, comparable sales, title verification, disclosure parsing, risk screening

The reasoning kernel stays stable across every domain. What changes is the playbook: which taxonomy applies, which sources are authoritative, what contradictions look like in that domain, what confidence thresholds matter, and what the output needs to say. Adding a vertical is a configuration exercise — taxonomy, source priorities, validation rules — not an engineering rebuild.

## The Structured Decision Record

Bernard's core output. A structured decision record contains:

| Component | Description |
|-----------|-------------|
| Evidence | Source documents preserved and linked |
| Claims | Typed assertions derived from evidence |
| Contradictions | Explicit conflict records with resolution paths |
| Confidence | Per-claim scores, decomposable and calibrated |
| Audit trail | Every extraction, verification, and resolution logged and timestamped |
| Views | Memo, packet, audit, agent payload — rendered from one record |

The record is readable by your team and by the agents that work alongside them. It does not expire. It is exportable. Your data is not used to train models, shared with other accounts, or made accessible outside your organization.

## Key Differentiators

| Dimension | Current Workflow | Standalone LLM | Bernard |
|-----------|-----------------|----------------|---------|
| Output | Folders of documents | Paragraph with footnotes | Structured decision record with linked evidence |
| Contradictions | Whoever noticed last | Smoothed into prose | Explicit conflict record |
| Confidence | Gut feeling | Implied by tone | Scored per claim, decomposable |
| Six months later | Lost in a shared drive | Session expired | Full record — reasoning and confidence intact |
| Audit trail | Reconstruct when asked | "Source: document.pdf" | Automatic, exportable |
| Automation ceiling | One task, one person | 1-2 steps, then human absorbs risk | Structured gates enable multi-step agent workflows |

## Company

- **Name:** Bernard, Inc.
- **Entity:** Delaware corporation
- **Headquarters:** New York
- **Founded:** 2025
- **Founders:** Timothy Kompanchenko (14 years in art market transaction infrastructure), Kiki Tovey (24 years in luxury hospitality and real estate operations)
- **Program:** Anthropic Startup Program
- **Status:** Limited early access
- **Contact:** support@hellobernard.ai

## Pages

- [Homepage](https://hellobernard.ai/): Three animated scenarios showing Bernard in action across fine art, real estate, and collector vehicles
- [How It Works](https://hellobernard.ai/how): 15 specialized AI agents coordinating research, verification, and risk screening
- [Why Bernard](https://hellobernard.ai/different): The structured decision record — evidence-linked, contradiction-tracked, confidence-scored
- [Team](https://hellobernard.ai/team): Founded by Timothy Kompanchenko and Kiki Tovey
- [Ukraine](https://hellobernard.ai/ukraine): Bernard's origins in preserving Ukraine's looted cultural heritage
- [Request Early Access](https://hellobernard.ai/start): Contact form for early partners
- [Download](https://hellobernard.ai/download): Desktop application for macOS, Windows, and Linux
- [Legal](https://hellobernard.ai/legal): Terms of use and intellectual property notices
- [Privacy](https://hellobernard.ai/privacy): Data handling, collection practices, and user rights