Topic · GUS.ai across the entire plant
A fleet of agents.
One plant.
Year three vision. Per-area coordinators, per-skid specialists, plant-wide cross-cutting agents.
Chapter 15
The
fleet.
The year-three vision. How a single-asset deployment scales to plant-wide orchestration.
The hierarchy
Four tiers.
One conversation.
Operator asks a question at any level. The agent at that level pulls down from its tier or escalates up. Same UI, same audit, same evidence discipline as the single-asset deployment.
Tier 1 · Orchestrator
One plant-level agent owns end-to-end batch flow — milk in, cheese and whey out. Watches throughput, yield, OEE, energy per pound. Brokers handoffs between areas.
Tier 2 · Coordinators
One agent per process area: raw, separation, make, whey. Owns recipe sequencing and per-area KPIs. Reports up to the orchestrator.
Tier 3 · Specialists
Per-skid agents — pumps, CIP circuits, HTST trains, membranes. Knows its equipment OEM-deep. Surfaces drift, suggests cleaning windows, drafts shift handoffs.
Tier 4 · Cross-cutting
Plant-wide specialists that span every area: alarm correlator, sanitation coordinator, energy auditor, lab traceback, recipe versioner, customer-complaint investigator.
Honest read
This is year three.
Not year one.
The single-asset deployment is production-grade today. Omni describes the multi-tier architecture that emerges as more rituals restructure.
- Tier 1 and Tier 4 are aspirational until enough of Tier 2 and Tier 3 exist to feed them.
- Each per-skid agent costs real ingest and prompt work. We add them in priority order, not all at once.
- Read-only stays read-only at every tier. Writes are not on the roadmap.
End of topic · GUS.ai Omni