In simple words: Reasoning systems organise steps; the output still needs verification.
What the phrase means here
The course notes use “reasoning architectures” for systems that analyse a problem, plan steps, and reach a structured answer. This is a broad phrase rather than one specific neural network like a CNN.
A practical view
A system may decompose a business question, gather relevant facts, compare options, and check whether the conclusion follows. A fluent explanation can still be wrong, so important claims need evidence and human review.
See the flow
question→facts→steps→answer→check
Simple example
For a stock shortage, the system might check demand, supplier delays, and inventory before suggesting actions. A manager verifies the recommendation.
Remember: Reasoning systems organise steps; the output still needs verification.