What if biology explained AI better than computer science? Behind each artificial intelligence project lies an often ignored reality: companies are less perfectly functioning machines.
Biology (more than computer science) is perhaps the most relevant reading framework to understand why transformations linked to AI are both so difficult and so essential. The problem isn’t the technology: it’s that we’re automating a story shaped by evolution.
Evolution does not seek perfection
Evolution is not an engineer. It doesn’t draw plans, run regression tests, or refactor legacy code. It only optimizes three things: survival, local adaptation, and short-term advantage.
Sound familiar? It should.
This is exactly how most organizations evolve their technology landscape. And it’s this idea that strikes me every time.
A workaround appears because a team needed to move quickly before the end of the quarter. An Excel file is created because the central system does not know how to handle a particular case, and it seems riskier to delete it than to keep it. A local marketing automation solution is approved in 2020 and miraculously survives two reorganizations… then the AI revolution. Not because anyone defended it, but simply because no one really took the risk of questioning it.
All of these systems are what biology calls vestigial structures.
In the living world, a vestigial structure is an anatomical remnant that once had a clear function but remains long after its environment has changed. The appendix, wisdom teeth or even the coccyx are examples. They are not defective; they are perfectly preserved echoes of a context that no longer exists.
Organizations work in exactly the same way.
Vestigial structures can be found in almost every business, provided you know where to look:
- Reports that no one reads, produced by systems that no one dares to turn off;
- Duplicated applications because neither team trusts the other’s data;
- Manual reconciliations inherited from previous financial crises prior to the current CFO;
- Governance committees of which even the members struggle to precisely explain the role;
- Validation circuits which remain mainly because their suppression seems politically risky;
- “Temporary” solutions which discreetly celebrate their eleventh anniversary.
The reason is simple: most businesses are not magnificent, methodically designed machines. These are successive layers of decisions accumulated over time, held together by a few heroic people armed with their spreadsheets.
No one understands dependency chains more completely.
This is why enterprise architecture often resembles an archaeological discipline more than an engineering discipline.
You can piece together the history of a company—its mergers, its crises, its strategic shifts, or its moments of institutional panic—simply by analyzing the naming conventions of Excel files, duplicate customer IDs, or the silent logic of some mysterious SharePoint folder.
With AI, we are not automating a designed system. We automate an accumulated story.
This is when the biological metaphor ceases to be a simple intellectual curiosity and becomes a strategic warning.
Many AI transformation programs are based on the idea that they add a layer of intelligence to fully understood, documented and controlled processes.
The automation logic seems clear. Expected efficiency gains are modeled. The business case is validated.
But the real foundation on which AI will run, learn and operate is not a rationally designed system.
It is a system shaped by evolution.
It contains workarounds, the origin of which no employee can precisely explain. It contains data influenced by decisions made by people who have long left the company. It contains informal knowledge that resides exclusively in Fred’s head, without ever having been documented.
The organizations that are most successful in their AI transformation are not necessarily those with the cleanest architectures.
They are the ones who have the most honest understanding of their own organizational evolution.
They carried out the archeology work. They know where the vestigial structures are. They have accepted difficult conversations to distinguish what is real business logic from what is just the scar of past events.
What this means for technology leaders
Today, the most important skill for a technology leader is to understand the organism before operating on it.
It is not the mastery of model architectures or prompt engineering.
It’s a form of organizational paleontology: the ability to read the layers of decisions accumulated over time and understand what they actually reveal about how the company works, rather than how the company would like to believe it works.
Before your next AI deployment, I invite you to ask yourself three questions:
- Do we understand why this process has this form today, and not just what it does?
- Have we identified informal knowledge holders whose expertise is not documented in any system?
- Are we automating real value creation or are we simply automating legacy complexity and making it work at machine speed?
I did not expect to find in a biology textbook the most relevant reading framework for understanding AI in business.
But evolution has always driven the largest transformation program in history.
And its most important lesson is not optimization.
It is the deep understanding of what we are actually working with.
