There was a time when knowing the latest framework was a competitive advantage. I used to know most of the J2EE spec by heart. If that sentence made you wince, you’re one of us.
But AI learns new frameworks faster than any human now. That’s no longer an edge.
What’s always been valuable — and will remain so — is the ability to see and operate within complex, interconnected systems. Systems theory. Remember PID controllers from school? That’s the foundation.
What does systems thinking actually mean?
It’s the ability to understand:
- Which elements are connected to each other — and how
- What actually causes the consequences you’re seeing
- Where feedback loops emerge — and whether they amplify or dampen behavior
Without this, you fall into the trap of local optimization. You speed up one component and accidentally break three others. Or, as I like to say, a random woodpecker shows up and takes down your entire system.
Amazon’s one-day delivery: a $40-50B feedback loop
Amazon has always been exceptional at calculating systemic consequences.
When they launched one-day delivery in 2019, it cost a staggering amount to implement. Plenty of people — internally and externally — asked: “Why bother?”
Here’s why:
- Faster delivery → more spontaneous orders
- More orders → more accurate demand forecasting per region
- Better forecasts → more efficient warehouse placement closer to customers
- Closer inventory → cheaper and faster delivery
This is a reinforcing feedback loop. Each step amplifies the next.
By estimates, this single initiative added $40-50 billion in annual revenue. They invested once. The loop keeps compounding.
At AWS, the same systems approach — combined with 5 Whys — helps build reliable services by managing bottlenecks and breaking destructive feedback loops between services when stability degrades.
Why systems thinking is more critical now than ever
Now, an attentive reader might say: “Misha, last week you wrote that only speed matters, and the week before that architecture doesn’t matter much anymore. Just build things! What feedback loops?”
Fair point. But here’s the thing — it’s the speed itself that makes systems thinking more important:
1. Speed of change = speed of risk
Every change is a risk. Code review won’t catch the moment you 10x the traffic to a downstream service. Or when an AI agent restructures the interaction between several services, and you only discover what actually changed under production load.
2. AI creates unpredictable feedback loops
AI models are already complex systems in their own right. You load data into them. You add system prompts. And you can change all of this orders of magnitude faster than traditional ML models.
AI-driven automation will create many diverse and poorly predictable feedback loops in your systems. If you don’t see those loops forming, you can’t manage them.
3. AI is great at local optimization — but terrible at system-level improvement
AI already produces good results at optimizing individual components. But for any serious improvement that touches multiple parts of a system? You still need a human who sees the whole picture.
With the current intensity of change, the systemic approach is the only path to meaningful improvement.
How to develop systems thinking
- Draw dynamic maps of your systems. Not just static architecture diagrams, but representations of what influences what and how: flows, dependencies, feedback loops.
- Study other people’s platforms. How does AWS retain millions of customers? How does Netflix manage traffic and determine which recommendations to show you? What loops are at work?
- Train specific thinking patterns: second-order thinking, bottlenecks vs. constraints, reinforcing vs. balancing feedback loops.
The bottom line
Want to stay relevant in the age of AI? Don’t just understand how individual parts work. Understand how the whole system behaves.
Frameworks will change. Languages will change. The ability to see the system — to find the leverage point, to predict the second-order effect, to design the feedback loop that compounds — that’s the skill that stays.
This is the foundation of how I approach every organization I work with. Not “fix the code” or “add more process.” Find the system. Find the constraint. Fix what matters.