AI agent
Software that uses AI models to complete tasks autonomously — reading inputs, making decisions within guardrails, and taking actions in your tools (CRM, inbox, ERP) rather than just answering questions.
Agentic AI
AI systems designed to pursue goals across multiple steps and tools with limited supervision — the difference between a chatbot that answers and an agent that resolves.
Workflow automation
Connecting your applications so data and tasks flow without manual work, typically via platforms like n8n or Zapier; AI adds judgment steps that rules alone cannot handle.
LLM (large language model)
The AI models that read and generate text — the engines behind agents. Different models excel at different tasks, which is why Omnifys routes across 15+ of them.
LLM routing
Automatically choosing the best model per task for accuracy, speed and cost, instead of locking everything to a single vendor.
RAG (retrieval-augmented generation)
Grounding an AI's answers in your own documents and data, retrieved at question time — so responses cite your reality, not the model's memory.
Human-in-the-loop
A guardrail pattern where the agent drafts or recommends, and a person approves before sensitive actions execute. Standard in our regulated-industry deployments.
Guardrails
The rules that bound what an agent may do alone: confidence thresholds, approval steps, forbidden actions, escalation paths.
Model drift
The gradual degradation of AI accuracy as the world (or a model version) changes. Continuous monitoring — included in every Omnifys subscription — catches it early.
CX automation
Automating customer experience touchpoints — support, order status, returns, follow-ups — so customers get instant answers and staff handle only what needs a human.