What Is Agentic Automation?

Agentic automation runs automation through AI agents that make decisions against a goal, rather than following steps written in advance. Classic automation says "carry out these steps". In agent-based automation you say "achieve this", and the agent decides some of the steps itself.

How does it differ from RPA?

RPA applies rules. The steps are defined in advance and the robot follows the same path the same way every time. An agent is given a goal and a set of tools, and decides for itself which tool to use and when.

That flexibility is not free. A rule-based robot's output is predictable and auditable. An agent's output varies.

When an agent, when a robot?

If the rule can be written, use a robot. It is cheaper, faster and easier to audit. The place for an agent is where the rule cannot be written: interpreting free text, extracting information from documents of varying format, situations with more than one possible path.

In practice the two work together. The agent decides, the robot does the work.

Governance is set up from the start

When a system that makes decisions goes live, three things have to be defined beforehand: the record of the decision, the route to reverse it, and the circumstances under which human approval is required. Without those three, agent-based automation does not get past pilot in an enterprise setting.

Our approach on this side is described on our agentic automation page.