Engineering and governance notes from Salvoc
Salvoc publishes practical resources — technical guides, architecture notes, implementation patterns, governance notes, and decision guides — grounded in the work of building AI systems and automation that hold up in day-to-day operation.
- Practical guardrails for AI systems in business operations
An AI system in business operations should be controlled, observable, and reversible — with the control set proportional to the consequence and the authority granted. This note answers what practical guardrails should surround such a system: input boundaries, output boundaries, tool and action permissions, human control, reversibility, observability, cost limits, failure handling, and change control. Capability is not authorization — the objective is the minimum sufficient control set for what the system may do, not maximum automation and not maximum control.
- What business systems include — and how to stop patching tools together
Business systems are the operating whole that connects workflows, software, data and controls. When the pieces are patched together without structure, the company still has a system — an unmanaged one. This guide explains what a business system actually is, what the common symptoms of fragmentation look like, and when integration, workflow automation, custom development, or a broader intervention is justified — including the honest conclusion that no new system is required.
- When custom software actually makes sense: a practical decision framework
Most software needs are better met by a standard product, a configured platform, or an integration of existing systems. Custom software is justified only when the operational fit genuinely requires it. This guide gives a structured way to decide — including the conclusion that building is not justified.
- What to automate first: a workflow automation prioritisation guide
Workflow automation is not the automatic answer. The workflows worth automating first repeat with a known rhythm, move defined inputs to defined outputs, follow rules a person can state, and hand work between systems in structured ways. This guide gives a practical prioritisation framework — including the valid outcomes that a workflow should be integrated, simplified first, kept manual, or deferred, and that no automation is required at all.