Trust Center

Model System Cards

Every Octopus AFI capability has a published system card. The cards use an identical structure so they can be compared and reviewed side by side.

Each card states what the AFI capability is for, what it must not be used for, which data it can read, where deterministic code ends and language begins, how it is evaluated, where it is limited, who approves its actions, and what is logged.

The cards contain no benchmark claims. They describe behaviour and boundaries.

Published cards

AFI capabilityScope
Variance Commentary AFI capabilityThis AFI capability detects budget-vs-actual variances, requests commentary from the responsible budget owner, and records the response as audit evidence.
Forecasting AFI capabilityThis AFI capability rebases rolling forecasts when source data changes and documents each adjustment.
Knowledge Q&A AFI capabilityThis AFI capability answers plain-language finance questions from the customer's own reconciled data and logged commentary.

Shared card structure

  • Purpose and intended use.
  • Out-of-scope uses.
  • Data sources and access scope, including read-only access and role-based access control.
  • Computation boundary: what runs as deterministic code and what the language model handles.
  • Evaluation approach.
  • Known limitations.
  • Human approval gates.
  • Logging and lineage.
  • Escalation path.

Rules that apply to every card

  • No figure shown to a user is generated by a language model.
  • Every outbound message to a person is approval-gated by a named human.
  • Access to source systems is read-only and scoped per source.
  • Every action is logged with who, what, and when.
  • Every figure carries line-item lineage back to the general ledger.

Questions and answers

What is a model system card?

A published document describing an AI system's intended use, boundaries, data access, evaluation, limitations, and controls, so a reviewer can assess it without access to the source code.

Do all Octopus AFI capabilities follow the same structure?

Yes. Every card covers the same nine sections in the same order, so cards can be compared directly.

Do the cards contain performance benchmarks?

No. The cards describe behaviour, boundaries, and controls. Deployment results are reported separately.

Who should read these cards?

Internal audit, model risk, information security, and finance leaders assessing whether an AI system can be relied on for reported numbers.

Related pages