Explainer
What is agentic AI?
In plain language, without the hype — and without pretending the field is finished figuring itself out.
What “agentic” means
A regular chatbot answers your question and stops. An agent takes steps — it can pick a tool, run it, look at the result, and try again. Given a goal, it decides what to do next, checks its work, and keeps going until it’s done or it hits a wall.
“Agentic AI” is a loose term for systems built around that loop: a language model in the middle, tools on the outside, and some scaffolding that decides when to stop.
How it’s different from a chatbot
- A chatbot replies once. An agent acts, observes, and replies over many steps.
- A chatbot uses words. An agent uses tools — search, code, APIs, files.
- A chatbot is stateless by default. An agent tracks progress toward a goal.
- A chatbot can be wrong quietly. An agent can be wrong expensively — so evaluation and oversight matter more.
What people are building
Research assistants that read and summarize dozens of sources. Coding agents that open pull requests. Ops agents that triage inbound tickets. Personal agents that book, plan, and draft. Robots that follow multi-step instructions in the physical world.
Most useful agents today are narrow — they do one job well, with clear guardrails and a human in the loop for the risky parts. That’s a feature, not a limitation.
What we do at meetups
Short demos of working things (however scrappy). Deep-dives on a paper or a repo. Q&A with someone who’s further along than you. Workshops when there’s appetite. Honest post-mortems when something didn’t work.
You do not need to be an engineer to be useful in the room. Builders, students, operators, educators, and curious neighbours all belong.