Library / Systems Engineering Principles Framework
Jump to passage
In this reading

Link to current text

Published source confirmed at last check

Source changed 2026-10-03 08:25:59 UTC · snapshot created 2026-10-03 08:26:43 UTC · last check 2026-10-03 10:10:13 UTC

2. Construct complete obtaining arrangements

GreenHeat-4’s managers initially propose buying a controller or having an AI Agent write one. The engineering team applies SYSE.24 and rejects those two phrases as a usable option set. A purchase is one relation inside an arrangement; an AI Agent is one possible performer of named Work. The team uses the six arrangement prompts and eight common questions in SYSE.24 to construct four whole arrangements for the same required greenhouse-control result.

ArrangementAgents, Work, Methods, and meansEvidence, integration, support, capability, and exit
Ready controller plus integratorThe controller vendor’s engineering team supplies a configured controller. The integration company’s team performs sensor, actuator, network, and commissioning Work. GreenHeat-4’s engineering team maintains greenhouse requirements and accepts the result.The controller supports the stated field interface and can be commissioned in nine weeks. Greenhouse-specific failed-sensor behaviour, configuration export, and the supervisory interface remain unsupported. Vendor support is offered for three years; changing the control logic requires vendor access.
Commissioned custom controllerThe engineering provider’s team designs and integrates a custom controller and supplies source, configuration, test evidence, and support.The provider estimates sixteen weeks against a twelve-week need date. It proposes source escrow but has supplied no representative failed-sensor evidence for this hardware family.
Internal development with AI assistanceGreenHeat-4’s controls engineers lead the Work. A general AI Agent assists with code and test generation. An independent controls specialist reviews safety-relevant behaviour.The company would retain knowledge and change access, but it has no qualified hardware-in-the-loop environment and no evidence that the team can finish assurance within twenty weeks. AI-produced code supplies no capability or acceptance result by itself.
Ready controller plus internal supervisory layerThe vendor controller retains local safety interlocks and manual fallback. GreenHeat-4’s controls team develops greenhouse-specific supervisory optimization through a documented interface.The arrangement is estimated at eleven weeks and preserves internal change capability above the safety boundary. It still depends on configuration export, interface timing, and the vendor update policy; failure of any one defeats the arrangement.

The Finance result compares resource and cash consequences. The Organization Change result states the proposed human–AI–provider Work allocation and its authority gaps. The safety result states the protected failed-sensor condition. The commercial and legal results state the proposed access, update, data, and remedy terms. SYSE.24 uses those results but does not recreate their Methods or decide their questions.