Nature is fascinating. Recently I learned about the term “stigmergy”. A natural process where individuals coordinate at a micro level to build complex systems that benefit them all. Ants, for example, don’t “design” their colonies top-down. They grow them through stigmergy.
Each ant makes small, local moves: dropping material, following a stronger trail, building where others already built. The environment stores those moves as signals that shape what happens next. Positive feedback reinforces what works; fading signals and changing conditions prevent the system from getting stuck. The result is a surprisingly coherent architecture that emerges from countless micro-interactions. It's fast and efficient coordination through traces.
So why doesn’t product design work like this?
Computer scientist Melvin Conway had a possible answer. In 1967, he introduced what later became known as Conway’s Law:
Organizations which design systems are constrained to produce designs which are copies of the communication structures of these organizations.
When organizations build around communication structures, they often create silos (Engineering, Design, PM, QA, etc.). Channels become tightly controlled, which makes it hard for signals to flow naturally between the people who are actually adding value. The impact is direct: iteration cycles slow down. We’re not talking about micro-interactions anymore, we’re talking about something closer to bureaucracy.
These structures also create another artifact that shouldn’t exist in a stigmergy-like process: handoffs. Handoffs add coordination overhead and cause information loss. Plenty of studies warn about the impact for organization: as handoffs become critical gates, and gates create tension between teams, this creates the need for more overhead costs.
But there are also human variables that make this harder. Ants operate like a single organism. Humans don’t. High efficiency in learning and creative teamwork depends heavily on psychological safety (Source). Without it, people share less, hide problems, and avoid honest feedback, exactly the kind of signal the system (or product) needs to improve. And if you look at the decline of many tech companies, there’s often a pattern: increasing pressure to perform ends up producing worse outcomes through the product.
Another variable, less popular to question lately, is the impact of data-driven organizations. Optimization is a powerful direction, but it can also create misdirection and bad incentives. Economist Charles Goodhart summarized this with Goodhart’s Law:
When a measure becomes a target, it ceases to be a good measure.
It’s all about what we treat as “good” reinforcement. If the reward is optimization itself, the design process can lose its purpose because the target stops representing real value and becomes an artificial game. That connects closely to machine learning: if you define the reward function poorly, you get “successful” behavior that’s misaligned with what you actually wanted.
So to try to answer my initial question: why are we still designing products this way?
My theory is that the tech industry hasn’t had strong enough incentives to change. If you look at design processes over the last decades, not much has changed at the core. We still talk about agile and lean design, but we operate inside organizations shaped around old communication channels, bureaucratic handoffs, and environments where psychological safety isn’t treated as a priority.
Now we’re entering another technological wave with AI. My bet is that the incentives will finally show up because products will become more dynamic, competition will get tighter, and the cost of slow iteration will be higher than ever. Some companies will be forced to redesign not just their products, but their whole organizational structure so they can finally evolve their process of designing.
Which company will do it first?