AI & LLMs
What is agentic AI?
Agentic AI describes AI systems that pursue goals through their own sequence of actions: planning steps, using tools, observing results, and adjusting, rather than producing a single response to a single request.
New to Bolt? Sign up free.
Technical definition
Agentic AI refers to systems built around a language model that can plan, act, and adapt over multiple steps toward an objective. Four capabilities define the pattern:
- Planning: decomposing a goal into an ordered set of tasks
- Tool use: calling external functions, APIs, databases, or a browser to gather information or cause effects
- Memory: carrying state across steps so later actions build on earlier ones
- Iteration: observing the outcome of an action and revising the approach
A chatbot answers. An agent acts, checks, and acts again.
In plain terms
Regular AI answers your question. Agentic AI goes and does the thing.
Ask regular AI how to check whether 40 suppliers raised their prices, and it tells you how. Ask an agent, and it opens the records, compares them, and hands you the list.
Why it matters
The shift is from advice to work. That is what makes agentic AI worth a small business owner's attention, because advice was never the bottleneck.
The practical version, for a company with no engineering team: an agent that reads inbound email and drafts quotes from your pricing rules. One that reconciles invoices against deliveries and flags gaps. One that builds an entire application from a description.
Two risks come with the capability. An agent that can act can act wrongly, at speed, so the permissions you grant it define the worst case. And an agent that reads outside content can be redirected through prompt injection. The design answer is the same in both cases: narrow permissions, human confirmation on anything destructive, and validation on the server rather than in the browser.
