Agentic AI
AI systems that take actions in the world over multiple steps, not just produce a single response.
In detail
Agentic AI describes AI systems designed to set sub-goals, take actions, observe results and iterate without prompting on every step. A chatbot answers a question. An agent books a meeting, sends a follow-up, creates a record in your CRM and adds a task in your project tool. Modern agents are typically built on a foundation model (Claude, GPT, Gemini) plus tool-use APIs that connect to email, calendars, databases and other software via standards like the Model Context Protocol.
Why it matters for Australian business
Gartner predicts agentic AI will autonomously resolve 80% of common customer service issues by 2029. For Australian SMBs the opportunity is automating routine workflows that currently consume staff time. The risk is that agents can make irreversible mistakes (sending wrong emails, deleting records) at machine speed, so the design needs human-in-the-loop checkpoints, audit logging and a hard limit on what the agent can change.