FindeckOS
FindeckOS LLC Legal ArtifactVersion: 2026-11-v1.1 • Effective: November 12, 2026

AI Use Disclosure

Key Summary Commitments

  • Finance Desk agents run on Claude by Anthropic under Anthropic's Commercial Terms, which do not allow Anthropic to train its models on your data.
  • AI output can be inaccurate, incomplete or outdated. Do not rely on it without human financial review led by specialized FindeckOS personnel and the executive team.
  • Agents cannot move money, pay bills or change your books.

1. Multi-Agent Architecture & Anthropic Commercial Terms

FindeckOS deploys six (6) specialized AI agents: (i) Bookkeeping Review Agent, (ii) Cash and Liquidity Agent, (iii) Profit and Unit-Economics Agent, (iv) Reporting Agent, (v) Concierge Agent, and (vi) Compliance Agent (International).

These agents are powered by Claude foundation models operated by Anthropic PBC under Anthropic's Commercial Terms of Service. By contract, Anthropic is strictly prohibited from training its models on your financial inputs, prompts, or generated outputs.

2. Technical Capability Boundaries (No Write or Execution Access)

FindeckOS AI agents operate in a strictly sandboxed, read-only analytical capacity. Agents CANNOT initiate bank transfers, move funds, pay bills, execute payroll, alter journal entries, modify your Chart of Accounts, or write data back to your source accounting software.

3. Probabilistic Nature of AI & Mandatory Human-in-the-Loop Review

Large language models and automated statistical agents are probabilistic systems. AI output may occasionally contain calculation variances, misclassifications of ambiguous ledger memos, incomplete regulatory checklists, or outdated assumptions regarding US GAAP, IFRS, or local compliance (such as SYSCOHADA).

You must not rely solely on unverified AI agent output for tax filings, debt covenants, capital allocation, or external investor disclosures without human financial review led by specialized FindeckOS personnel and the executive team.

Executive Business Plan models and narratives are first drafted by an AI agent from the information and files you provide and from your answers on consultation calls. Every number is an estimate tied to a stated assumption, and the output is labeled a draft until a FindeckOS CFO reviews and approves it. Do not present a draft plan to a lender or investor before that review.

4. Escalation Governance

The Concierge Agent is programmed to answer routine factual queries from your synced metrics. Whenever a query involves strategic judgment, financing structure, or accounting policy interpretation, the system flags and routes the inquiry to specialized FindeckOS personnel and the executive team for human follow-up or a scheduled advisory session.