Arresting and Visualizing Complex Conditional Logic

Principles first approach

Before subjecting and evaluating popular SaaS platforms and emerging AI agents for their value-add potential, we should first agree on the principles and best practices to use for those evaluations.

Best practices… yes, they exist

In 2000, Jesse James Garrett introduced a set of flow-diagramming best practices that reimagined electrical schematic conventions for interaction and information design. By assigning logic-based definitions to simple geometric shapes, he transformed them into a symbolic language. These symbols, when combined correctly and consistently, allow information architects and designers to articulate complex, condition-heavy funnels with clarity and precision.

After auditing three SaaS platforms, I found that each used symbology inconsistently—and none provided a clear, comprehensive definition of what each geometric shape represented or how it related to a logical operation. Simply labeling a trapezoid as a “trapezoid” offers little value to a designer (especially newbies) trying to use a trapezoid meaningfully in a diagram. Today’s SaaS platforms…

• …fail to give the geometric shapes contained in their libraries clear definition and purpose as they might relate to logical considerations
• …fail to encourage designers to consider flow diagrams as an alignment artifact versus a design artifact. If you’re adding a bunch of color to your diagrams you’re likely adding too much detail.
• …fail to establish their own internal best practices for use or embrace other established versions.

At present, AI models still struggle with logic and reasoning. They can produce flow diagram with some reliability, but asking them to communicate logic using a predefined symbology while adhering to best practices remains a fool’s errand. Many models can generate a flow diagram from a prompt, yet ask for the same diagram twice and you’ll likely get two different results. This inconsistency underscores a simple truth: for now, humans remain the more reliable producers of logical structure.

While getting better, today’s top AI models and agents still struggle with logic and reasoning. They can produce flow diagrams with some reliability, but asking them to communicate those same flows using a predefined symbology built upon (and adhering to) best practices remains a fool’s errand. Many models can generate a flow diagram from a prompt, yet ask for the same diagram twice and you’ll likely get a different result. This inconsistency underscores a simple truth: for now, humans remain the more reliable producers of logical flow diagrams and structures.