Watch: Agentic AI — workflows vs. agents

Before you decide that a task needs an agent, it helps to see what a well-defined workflow can already do. This short video makes that boundary concrete.

Play

What to notice

As you watch, pay attention to three distinctions:

• A workflow follows a known path with steps that are mostly decided in advance.
• An agent chooses what to do next based on the goal, context, and results it observes.
• A useful system can combine both: keep the predictable parts fixed and give the uncertain step room to adapt.

The practical lesson is simple: use the smallest amount of autonomy that makes the task work.

How this connects to skills

The video focuses on workflows and agents. A skill fits alongside them as the reusable method that explains how one kind of work should be done.

For example:

• the workflow might collect rough notes and prepare a draft;
• the agent might decide which notes need clarification;
• the skill might provide the editorial method and quality checklist.

The companion article explains that relationship in more detail. This video gives you the visual starting point.

Try this after watching

Choose one task you do repeatedly and write down its path. Mark each step as one of these:

1. Fixed: the same action happens every time.
2. Method: the work needs a repeatable standard.
3. Decision: the next action depends on what the system finds.

Fixed steps belong in a workflow. A repeatable method belongs in a skill. Decisions are where agent behaviour may earn its place.