Skip to main content

Thinking in Systems

Most AI tools analyze problems in isolation. Ask about automating customer support and you get an answer about customer support. You do not get the second-order effect on employee morale, the third-order effect on company culture, or the feedback loop where cost savings lead to worse service which leads to customer churn which eliminates the savings. This chapter trains you to see interconnections that AI tools consistently miss.

Why This Matters: James and the Three-Line Analysis​

James had finished reading the scenario brief. A bank replacing its loan officers with AI. He pulled up a blank document and started typing his analysis. "Revenue impact: lower salaries. Customer impact: faster loan processing. Competitor impact: pressure to follow suit." He leaned back. "Three domains, three effects. Done."

Emma glanced at the screen. "What happens to the loan officers?"

"They get laid off. That's the employee impact. Four domains."

"And then what?"

James frowned. "What do you mean, 'and then what?' They find other jobs. The bank saves money. That's the whole point of the decision."

"When a bank fires its loan officers in a town where those officers coach Little League and sit on the school board, what happens to the bank's reputation in that community?"

"That's..." James paused. "Okay, that's a different kind of effect. But it's speculative. I can't map every hypothetical consequence of every decision. I'd be here all day."

"At your old company, when procurement switched suppliers to save 12% on packaging, what happened?"

James went quiet. He remembered that one clearly. "The new supplier's boxes were slightly smaller. Products didn't fit the same way on retail shelves. Three chains dropped us from their displays. We lost more in shelf space than we saved on boxes."

"Did anyone in procurement see that coming?"

"No. Because packaging and retail display were different departments. Nobody traced the connection."

"That's what I'm asking you to do here. Not four domains with one effect each. Five domains with chains of consequences that connect across boundaries. Your procurement team saw a packaging problem. It was a shelf-space problem. It was a revenue problem. Every link in that chain was invisible until it broke."

James looked at his three-line analysis. It suddenly looked like a table of contents for a much longer document.

"I identified the problem and the solution," he said. "What else is there?"

"Everything your procurement team missed." Emma picked up her coffee. "Draw me a map. Start with the decision at the center. Branch out to five domains. For each domain, ask 'and then what?' at least three times. And look for the loops, the places where a downstream effect circles back and changes the original decision."

"Loops?"

"The bank saves money by cutting loan officers. Service gets worse because AI can't read a farmer's handshake the way a human can. Customers leave. Revenue drops. The savings disappear. That's a loop. The cost-cutting decision undermines itself."

James stared at the blank document. He'd been thinking in straight lines. Problems, solutions, outcomes. Each one isolated, each one terminal. Emma was describing something that curved back on itself.

"How many of these loops am I supposed to find?"

Emma was already at the door. "At least three. And I want mechanisms, not just labels. Tell me why each effect causes the next one. If you can't explain the why, the connection isn't real."

She paused. "I'll be back in an hour. Your map should look messy. If it's neat, you stopped too early."

She left. James sat with the scenario and his three-line analysis and the growing suspicion that the problem was not the bank's decision. The problem was how he was looking at it.

Exercise 1: The Cascade Map​

Layers Used: Layer 1 (Predict Before You Prompt), Layer 6 (Iterative Drafts)

James is staring at a blank document with a three-line analysis he no longer trusts. So are you.

Building On Previous Chapters

You will use the Error Taxonomy from Chapter 2, Exercise 1 to identify errors in causal reasoning. Systems errors build on the error detection skills you practiced there.

Draw Your Cascade Map (Before Touching AI)​

You receive a single decision. Without AI, draw a cascade map on paper or in a document, tracing effects across at least five domains: employees, customers, competitors, regulators, and the organization's own internal knowledge base. Identify at least three feedback loops (where an effect circles back to amplify or dampen the original decision). This map is your Draft 1, submitted before any AI is consulted.

Choose Your Scenario​

Scenario A (Finance): "A major bank decides to replace all loan officers with AI agents."

Choose one. The exercises work identically regardless of which you pick.

Your Deliverable

A cascade map (hand-drawn scan or digital document) showing: the central decision, at least 5 domains affected, first-order effects in each domain, at least 3 second-order effects, at least 3 third-order effects, and at least 3 feedback loops clearly labeled (e.g., "cost savings leads to reduced service quality leads to customer churn leads to reduced revenue leads to negated cost savings"). Each effect should have a one-sentence explanation of the mechanism.

1Your Work

I am a student learning systems thinking. Before using AI, I created a cascade map tracing consequences of a decision across five domains with feedback loops.

The scenario I chose:

Please: (1) Evaluate the completeness of my map -- which important effects or domains did I miss? (2) Rate each of my feedback loops: are they logically sound? Would they actually occur? (3) Identify at least 3 second or third-order effects I missed that are non-obvious but important. (4) Rate the overall sophistication of my systems thinking from Beginner / Developing / Proficient / Advanced. (5) Do any of my causal chains have logical errors -- effects that would not actually follow from the cause I described?

Here is my cascade map:

Finally, complete the Thinking Score Card for this exercise: Independent Thinking (1-10), Critical Evaluation (1-10), Reasoning Depth (1-10), Originality (1-10), Self-Awareness (1-10). For each score, give a one-sentence justification.

2Get Your Score

Discuss with an AI. Question your scores.
Come back when you have your BEST evaluation.

Deliverable Template (click to expand)

CASCADE MAP TEMPLATE

  • Central Decision: ___
  • DOMAIN 1 [Employees]:
    • 1st-order effect: ___
    • 2nd-order: ___
    • 3rd-order: ___
  • DOMAIN 2 [Customers]:
    • 1st-order: ___
    • 2nd-order: ___
  • DOMAIN 3 [Competitors]:
    • 1st-order: ___
    • 2nd-order: ___
  • DOMAIN 4 [Regulators]:
    • 1st-order: ___
    • 2nd-order: ___
  • DOMAIN 5 [Internal Knowledge]:
    • 1st-order: ___
    • 2nd-order: ___
  • FEEDBACK LOOP 1: [A] leads to [B] leads to [C] leads back to [A] | Type: Amplifying/Dampening | Mechanism: ___
  • FEEDBACK LOOP 2: ___
  • FEEDBACK LOOP 3: ___

What Happened With James​

James stared at his map. It sprawled across the page, arrows crossing between domains he'd initially treated as separate columns. The employee domain connected to the customer domain through community trust. The competitor domain connected to regulators through industry precedent. His neat four-line analysis from an hour ago was buried under a web of connections he hadn't seen coming.

The feedback loops surprised him most. He'd found one where cost savings led to service degradation led to customer attrition led to revenue loss that erased the savings. It was the packaging-and-shelves problem from his old company, just wearing a different suit.

"The loops are the part I would have missed completely," he told Emma when she returned. "I was thinking in straight lines. Problem, solution, done. But the solution creates new problems, and some of those problems circle back and undo the solution."

"How does that change the way you'd analyze the next decision you see?"

James thought about it. "I'd stop looking for the answer and start looking for the second thing that happens. And the third. And whether any of them come back around."

"That's the muscle this exercise is building. Not the map itself. The habit of asking 'and then what?' one more time than feels necessary."

The Lesson Learned​

Linear analysis treats each domain as its own container. Cascade mapping forces you to trace effects across those boundaries and find the loops where consequences circle back to reshape the original decision. The skill is not the map. The skill is the habit of asking "and then what?" one more time than feels necessary, until you find the connections that nobody in the room was looking for.

Flashcards Study Aid​