AI agent
A software system that uses an LLM to execute a workflow against the customer's connected systems, reading from sources, taking actions, and writing results back through the standard authorization model.
In more detail
The agent extends the LLM with three capabilities the chatbot does not have. It reads from the customer's systems through the standard authorization model. It runs a chain of steps that compose into a finished outcome rather than a single message. It writes results back to the source system on the customer's approval policy.
The architecture separates reasoning (the LLM) from action (the integration adapters), with the workflow engine in the middle orchestrating the chain. The audit log captures every action.
Where this shows up at Ceven
Ceven's agents are specialized: the finance agent owns close, the HR agent owns lifecycle, the IT agent owns JML, the procurement agent owns sourcing. They share a common substrate (audit, security, integration adapters, voice). The customer interacts with workflows rather than individual agents.