Layer 02 Agentic Intelligence

Hold and rebook

Pausing a delivery that cannot leave on time, usually because the store or warehouse is not ready, then cancelling the courier booked for it and booking capacity again for when the order will be ready. Done by hand it means several systems and a phone call per order. In Nash, an agent watching store readiness proposes the hold, an operator approves it on the route in Execute, and the rebook runs from there.

Where this fits

Layer 02 of the Nash platform: the agents that observe, analyze, and act. Ask Nash for natural-language questions, custom workflow agents triggered on real events, durable agent memory across runs. Where the system makes decisions about decisions.

Related terms

Agents: Autonomous actors that pursue goals you set, e.g. keep SLA above 95% or re-dispatch any order unaccepted after 4 minutes. Pre-built or operator-configured. Scope is explicit: an agent reads the data you have connected, acts only inside the guardrails you define, and escalates anything outside them to a person. By default, actions that reach outside the system, such as cancelling an order, calling a customer or issuing a refund, wait for an operator to approve them on the route the dispatcher already works.
Agentic AI: AI that decides and acts on its own within defined guardrails, not just recommends. In logistics, an agentic system evaluates conditions and executes, rerouting a load or resolving an exception, rather than surfacing a suggestion for a person to approve. Agentic describes how a decision gets made; autonomic describes a system that runs continuously. Nash is autonomic, and agentic action is one mechanism inside it. Guardrails are the boundary, not a suggestion: what an agent may read, what it may change on its own, and what it must hand to a person are all configured before it runs.
Workflows: Order-level routing logic. Decides which dispatch strategy applies, which modules are invoked, what happens when conditions change.
Memory: The learning substrate. Provider scores update, routing improves, promise accuracy tightens with every outcome. Skills turn operational know-how into reusable playbooks.
Operational alerting: The signal an operator gets the moment something needs attention: a delivery running late, a provider that just canceled, an ETA sliding past its window. Nash's agents raise it against live context and route it to whoever can act, so problems surface before a customer notices.