AMR lifecycle economics: a reproducible fictional stress test

A clone of the AMR preset with all failure switches disabled, stage set to Scale, and descriptive scenario identity changed. This is an unperturbed hypothetical comparator, not observed operations or the original preset's enabled-failure mix.

The baseline and five stress variants use the same fleet, horizon, prices and operating assumptions. Selected failure switches and their declared severity change; descriptive scenario identity changes so each result is distinguishable. Severity is not an observed probability.

• Severity is a scenario input to versioned consequence formulas, not an empirical probability or measured effect size.
• Only the declared failure settings (enabled state and selected severity) differ between comparable runs, apart from descriptive scenario identity. Inputs, results and hashes are included for independent recalculation.
• Productive value is an assumed customer benefit; contract revenue is shown separately and is not added to that benefit. Lifecycle net value is not supplier profit, NPV or a forecast.
• The existing engine does not discount cash flows or model tax, financing, utilization distributions or correlated real-world failure probabilities.
• The illustrative exposure counts in Product System context are held constant; the failure engine does not predict accepted pallet transfers or operator staffing. Do not read the cost-per-declared-success diagnostic as stress-adjusted throughput economics.
• A delayed supplier may leave modeled release weeks unchanged when the declared roadmap already contains slack. Read the changed lead-time field rather than claiming every stress must move every output.
• No scenario in this table reverses the result, so it shows sensitivity, not the point at which the investment should stop. In this model, lifecycle net value reaches zero when the productive value of an hour falls to the cost per productive hour of the run: 6.75 EUR in the baseline and up to 8.61 EUR under the declared stresses, against the assumed 34 EUR. A site-measured value below that level turns the modeled result negative; at that level it is zero. That is the stop condition for this expansion decision.
• Before investment, replace assumptions with scoped site evidence, capture omitted costs and compare alternatives. A favorable modeled signal still recommends experiment_only.

These files were prepared on 7 October 2026 before publication, with no individual human editorial approval recorded. Their status fields retain that preparation state; they do not grant permission or establish when this page became available. AI tools supported preparation. No customer results or release approval are established.

The file manifest binds the download bytes to their local sources. The downloadable JSON includes the inputs, limitations and provenance needed to inspect the result.

Payload SHA-256: 2a612c0a5b74741418b8a6c242970f483b0aba1cccd41d83a3875eab3315d959

Use the Physical AI Product System to connect claims, requirements, configuration, evidence and an accountable human decision.

Product Leadership Mission: Lead one product, platform, launch or operational capability to a named evidence-backed milestone.