MMOOZZGG.com

Glossary

What is agentic AI

In short

Agentic AI is an approach where a neural network receives a goal and carries the work to a result on its own. It plans the steps, calls software tools and checks what came out.

AI agents · agentic AI

MMOOZZGG.com editorsChecked: 2 min read

For a child

Imagine two building sets. The first has a picture, and you place each piece exactly as the picture shows. The second has no picture, only a task: build a bridge over the stream. You decide where to start, try, see if it holds and rebuild. Agentic AI is like the child with the second set: it gets a goal and finds the way itself.

In plain words

A usual AI model works like a conversation partner: you ask, it answers, and then you decide what to do next. In the agentic approach you give the model the whole task and access to the right programs. It splits the work into steps, does the first, looks at the result and chooses the next step. It continues until the goal is reached. For example, you ask it to compare three suppliers. The system finds their price lists and puts the prices in a table. When it notices a gap in the data, it asks for the missing part instead of stopping after the first answer.

For business

What matters to a company is not the term itself but how much of a process it can hand over without constant supervision. Agentic systems fit tasks with many branches: sorting incoming email, matching documents, preparing commercial offers. The cost depends on the number of steps and model calls per task, the cost of integrations and the time for quality checks. The main risk is that a mistake in an early step travels down the chain. So before launch, settle three questions: which actions the system takes by itself, where a person is needed and how you will learn about a failure. An AI solution map shows where this approach pays back and where simple automation is enough.

For a specialist

Agentic AI describes a class of systems in which a large language model with tools runs a multi-step task instead of only generating text. In the article “Building effective agents” (19 December 2024), Anthropic divides such systems into two types. Workflows connect the model and the tools through paths fixed in code. Agents choose their next steps and tools themselves, using feedback from the environment. The article describes common patterns: prompt chaining, routing, parallelization, orchestrator-workers and the evaluator-optimizer loop. In practice teams combine both forms: a fixed frame where predictability matters and free choice of steps where the path is unknown. The engineering questions are tool descriptions, a step budget, logging and evaluation on a set of scenarios.

Example

A purchasing team asks the system to prepare a request for materials. It checks stock in the accounting software, finds three suppliers, compares terms and prices and gives the manager a ready draft, with notes where data was missing.

Related terms

Questions and answers

What is the difference between agentic AI and an AI agent? An AI agent is a specific program that does a task. Agentic AI is wider: it names the approach itself and the whole class of systems where a model runs the work step by step, including teams of several agents.
Does a business always need an agentic approach? No. If the path of the task is known in advance, a fixed script where the model does single steps is more reliable and cheaper.

Give the model freedom of choice only where there are too many options to write down in advance.

Sources

Not sure whether your company needs this?

AI solutions map
Contact us