<!-- Canonical: https://myoctopus.ai/trust/system-cards/forecasting -->

1.  [Home](/)
2.  /[Trust](/trust)
3.  /[Model System Cards](/trust/system-cards)
4.  /Forecasting AFI capability

Model System Card

# Model System Card — Forecasting AFI capability

Version 1.0

Last updated September 2026

Public document

This AFI capability rebases rolling forecasts when source data changes and documents each adjustment.

All forecast arithmetic runs as deterministic code against a reconciled data layer.

## 1\. Purpose and intended use

*   Rebase a rolling forecast when general ledger, billing, or bank data changes.
*   Apply customer-defined driver logic and documented assumptions.
*   Produce scenario comparisons against a named baseline.
*   Record the reason for each adjustment alongside the adjusted figure.

## 2\. Out-of-scope uses

*   Replacing the customer's planning platform or acting as the system of record for the plan.
*   Setting targets, approving budgets, or committing spend.
*   Market, security, or price forecasting outside the customer's own financial data.
*   Any use where a forecast is presented as an audited or guaranteed outcome.

## 3\. Data sources and access scope

| Source | Access |
| --- | --- |
| General ledger and sub-ledgers | Read-only, scoped per source and per entity. |
| Planning platform | Read-only baseline plan and scenario versions. |
| Operational feeds | Billing, bank, and headcount data where the customer connects them, read-only. |
| Assumptions | Driver logic and assumptions authored and approved by the customer's finance team. |
| Access control | Role-based access control, scoped per source system. |

## 4\. Computation boundary

| Layer | Responsibility |
| --- | --- |
| Deterministic code | Aggregation, driver maths, currency conversion, phasing, scenario arithmetic, and reconciliation. |
| Language model | Explaining a movement in plain language and drafting the accompanying narrative. |
| Boundary rule | The same inputs and the same assumptions produce the same forecast figure every time. Model drift cannot change a number. |

## 5\. Evaluation approach

*   Backtesting against closed historical periods using only data available at the time.
*   Reconciliation of every forecast baseline to the general ledger before publication.
*   Assumption-change tracking: each rebase is compared against the prior version and the delta is attributed to a specific driver or input.
*   Determinism checks: repeated runs on identical inputs must return identical figures.

## 6\. Known limitations

*   Forecasts inherit the quality of the customer's driver logic and mappings.
*   Structural events such as acquisitions, disposals, or reorganisations require the customer to update the model structure first.
*   Sparse or short data history reduces the usefulness of driver-based projection.
*   A forecast is an estimate. It is not a commitment and should not be relied on as one.

## 7\. Human approval gates

*   A named human approves publication of a rebased forecast to a wider audience.
*   Assumption and driver changes are approval-gated and versioned.
*   Any outbound message referencing forecast figures follows the same approval gate as variance commentary.

## 8\. Logging and lineage

*   Every rebase is versioned with the inputs, assumptions, and approver recorded.
*   Every figure traces to the source line items it aggregates.
*   Adjustment reasons are stored with the adjustment, not in a separate document.
*   Version history is exportable for audit.

## 9\. Escalation path

*   Failed reconciliation blocks publication and raises a data-quality item.
*   A material unexplained movement is escalated to the customer's finance owner before distribution.
*   Availability incidents are published on the status page.

## Questions and answers

### Does the Forecasting AFI capability calculate figures with a language model?

No. Arithmetic, consolidation, and currency conversion run as deterministic code against a reconciled data layer. The language model handles wording only.

### What access does the Forecasting AFI capability have to source systems?

Read-only access, scoped per source system, with role-based access control applied before data is retrieved.

### Is human approval required before the Forecasting AFI capability acts?

Yes. Every outbound message to a person is approval-gated and a named human signs off before anything sends.

### What is logged when the Forecasting AFI capability runs?

Every action is logged with who, what, and when, and every figure carries line-item lineage back to the general ledger. Logs are exportable for audit.

This system card describes documented product behaviour and controls. It is not a warranty of outcomes. Figures produced by any AFI capability remain subject to the customer's own review and sign-off.

## Related pages

*   [All Model System Cards](/trust/system-cards)
*   [The Octopus AI Constitution](/trust/constitution)
*   [Transparency Report](/trust/transparency-report)
*   [AI governance in finance (SR 11-7, DORA)](/ai-governance-finance)
*   [Trust Center](/trust)
