As always the explanation of very basic concepts is very difficult, resulting often into a tautology. I will try here to formulate a clear explanation of what a workflow is and how it is used in KNIME.
A workflow is an analysis flow, that is the sequence of the analysis steps necessary to reach a given result. It is the pipeline of the analysis process, something like:
1. reading data
2. cleaning data
3. filtering data
4. training the model
Each step of the data analysis is executed by a node. The sequence of nodes makes the workflow.
In KNIME the workflow is represented graphically and each node is represented in the workflow as little icon. This graphical representation makes it easy to keep the overview of the analysis process and to document it.
In the KNIME whitepaper (http://www.kdd2006.com/docs/KDD06_Demo_13_Knime.pdf) a workflow is more formally defined as:
"Workflows in Knime are essentially graphs connecting nodes, or more formally, a direct acyclic graph (DAG). "
An easy example of a KNIME workflow is shown below.