Agentic Layer
AI that onboards, answers and drafts — inside code-enforced guardrails.
Three agents over the whole suite: an Admin Agent that onboards a warehouse by chat, a Warehouse Agent on every screen that answers with dated evidence, and a nightly Ops Agent that drafts tomorrow for a human to approve.
0
free-text guesses — every figure from a live query
Human
approves every plan, in every mode
How it works
Agentic Layer, in three moves.
The AI orchestrates and explains; it never does the maths. Every figure comes from a live query against the site’s own data, every write is staged as a diff for human approval, and the scope an agent can see is injected by the server — not the prompt.
01
Onboard
The Admin Agent walks a new site from identity to first run, probing each gate from reality.
02
Ask
The Warehouse Agent answers any operational question from live data — or says it cannot.
03
Draft
The Ops Agent prepares tomorrow overnight. A human approves before anything runs.
Capabilities
- Onboard by chat: create the warehouse, register data, walk the gates, fire the first run
- Page-aware agent with tools — no free-text guesses, every answer dated
- Nightly Ops Agent drafts crew, settings and risks: Approve / Adjust / Reject
- Server-injected scope, staged writes, human gates
- Grounded in the live capability registry and on-screen help
- Every agent interaction metered and logged