An AI agent is a software system that uses a language model as its reasoning engine to plan and execute multi-step actions against external tools — APIs, databases, file systems, browsers — rather than only producing text. The agent reads context, decides what to do, calls a tool, observes the result, and either continues, completes the task, or escalates.
An agent is defined by four components: a model (the reasoning core), a tool layer (the things it can do in the real world), a memory or state layer (so it can act across turns and sessions), and a policy or guardrail layer (so it does not do things it should not). Remove any of these and you no longer have an agent — you have a chatbot, a script, or a research demo.
The practical test: an AI agent can be measured by outcomes (bookings made, tickets resolved, invoices reconciled), not by messages sent. If the metric that matters is conversational, it is probably a chatbot. If the metric is operational, it is an agent.
How Nordix uses it
Nordix BIOS is a fleet of AI agents — one per business workflow (reservations, takeout, suppliers, staff Q&A) — each with its own tool set, policies, and outcome metrics. Agents share state through the tenant's data layer so a customer's reservation context is available to the takeout agent the next day. Every tool call is logged and reversible where the underlying system allows it.