Dashboard, AI, and review

The dashboard answers three executive questions: Is the project fundable? Is the value case credible? What must be true to proceed? Start with the selected scenario and funding gap, then inspect the cash low point, reserve breaches, break-even, NPV, ROI, payback, IRR, margin, downside spread, evidence quality, and readiness warnings.

Do not interpret green indicators as approval. Confirm that the horizon is sufficient, financing is genuinely available, line items are complete, assumptions are validated, and non-financial success criteria remain satisfied. Investigate any disagreement between the recorded recommendation and the deterministic signal.

Appropriate AI interaction

AI can suggest missing income or cost categories, draft assumption records, identify contradictions, propose downside tests, summarize evidence gaps, challenge a recommendation, and draft report language from the saved model. It should cite the model objects behind its observations and clearly disclose uncertainty. It must not invent organization-specific amounts, validate its own drafts, silently edit formulas, or authorize the project.

In Income & Costs, AI Suggest Items opens a review modal while the model is analyzed. The modal explains each proposed item, its classification, calculation basis, confidence, rationale, assumptions, and any unresolved numeric input. Accept or reject recommendations individually; the workbench changes only when Add accepted items is selected. Accepted items with missing numbers enter disabled and remain unusable in the forecast until validated.

In Scenarios, AI Suggest Scenarios follows the same controlled workflow. The modal explains each proposed stress test, income and expense multipliers, opening delay, assumptions, rationale, and valid line-item adjustments. It also calculates a preview of ending cash, funding gap, NPV, and months below reserve using the deterministic engine. Accept or reject scenarios individually; only Add accepted scenarios changes the model. Recalculate and revalidate the preview whenever base inputs change.

An explicit instruction such as “fill, populate, or complete this Project Viability cash flow from the following brief” builds and applies traceable AI-draft inputs to the active model. Questions that ask for an explanation or review do not auto-apply changes. Recurring annual operating quotes are converted to monthly inputs by the workbench; unknown quantities remain disabled and are reported as validation questions. Review every applied draft, especially demand, staffing, startup cost, timing, and financing assumptions.

Executive review record

Resolve or explicitly accept material warnings. Record go, conditional go, no-go, or defer; explain evidence and tradeoffs; name the most important risks; list measurable approval conditions and next actions with owners. A conditional go without conditions is incomplete. Preserve AI reviews as advisory snapshots, while the executive rationale remains the accountable decision record.