Why do most companies die so fast?

Why do most companies fail, and why do they fail faster today than sixty years ago? Why are founder-led companies performing better than other companies, and why do other human-created social constructs like cities live forever as opposed to companies??

I have been asking many questions about the future of enterprise software and the role AI will play, and it recently occurred to me that maybe I was asking the wrong questions. Maybe we needed to change our frame of reference on the role that software plays in company operations as we move paradigm from software as calculators to software as reasoning machines — but how?

I’m a sociologist by education, but have spent my whole life in tech; and never before have I seen a bigger intersection between the two fields than now. The rise of AI-based technologies will push us to fundamentally rethink many social and economic models, and the winners will be those who are either lucky enough or happen to design for that.?

In Beyond Work, we believe that most enterprise software will be irrelevant or anachronistic in 10 years, but what will we have instead? We bet the focus will be on the work, not the app. How will that manifest itself? What to build or design?

To start, we need to consider what work is. Work is a unique human construct, a group of people centered around creating an output or outcome with a common goal.

In a business, this outcome is typically a return for shareholders. However, for a mission-driven NGO it could be creating an impact for stakeholders. Either way, underlying most work is the idea of purpose or intent — and more than ever, we want work to mean something.?

Typically, companies have a command and control process, starting with a CEO, tiers below, etc. The top issues marching orders which are then replicated downstream until they ultimately turn into actions and outcomes. Originally, these processes were manual, using such advanced technologies as pen and paper and, later, telegraphs and phones. Finally, as we went through the digital revolution, these processes got digitized and turned into software, all in the name of efficiency.

So why, with our state-of-the-art digital business processes, are big companies dying faster than ever??

With the digital revolution, we may have lost something important. What if, by digitizing top-down processes, they accidentally became ossified, unable to change? While inefficient in some ways, prior methods of communicating had the advantage of being highly adaptable.?

Maybe losing the ability to adapt and quickly change is why the average lifespan of a Fortune 500 company today is 18 years vs. 61, just six decades ago. While we as a software industry have evolved and created new methods to be more agile, what if the software is the problem?

However, companies are not the only human construct to organize and enable outcomes for groups of people.? We have many others, including religions, governments, and one I personally find very interesting: cities.?

Cities are interesting as an organizational template, as they, in most cases, are immortal. Except for massive natural disasters, most cities live forever, and even man-made catastrophes like the nuclear bomb did not prevent Hiroshima and Nagasaki from becoming thriving cities again. Cities are distributed, bottom-up, culture driven, not command and control; their returns scale better over time than companies and use fewer resources (super and sublinear scaling).?

Much of my thinking on this topic is influenced heavily by Geoffrey West and the idea of emergent complex systems, starting with simple rules and then scaling up to hyper-complex systems. He encapsulates this in the brilliant book “Scale: The Universal Laws of Life, Growth, and Death in Organisms, Cities, and Companies” but one thing he doesn’t discuss as much is the role of software and digitization in how these systems perform. I think this might be where we can extract the biggest lessons for the future.?

Because of information entropy, command and control systems tend to perform poorly at scale. Too much information is lost unless the actions and outcomes are incredibly simple.?

Further, the intent and reasoning are lost as actions are encoded in software, making it harder to change and adapt. In that case, this information entropy worsens, and companies lose their ability to adapt to their environment, proving evolution right: survival is about adapting to be the fittest.

Ironically, we see the same problem mirrored with AI agents, as they have a hard time solving simple second or third-order problems. Again, the problem is information entropy.

For AI, this is visible in the Jarvis-1 paper, which shows the sharp drop-off in efficiency for any task beyond the first order (in this case, the open-world game Minecraft, which is still significantly simpler than the messiness of the real world).

We can laugh and use this as a case for why humans are still superior to AI in some ways, but what if we flip the conclusion? Instead, question if the problem is the command and control nature of the instructions for both agents and humans.?

When we play Minecraft, we are better than agents because we are having fun, but when it comes to work, we are not that much better than Jarvis-1, as the declining lifespan of most companies shows.?

Daniel Pink, arrived at similar conclusions in his book Drive 14 years ago when he pointed out we perform worse when we are externally directed than when we can be intrinsically motivated.

Why are decentralization and culture better drivers for longevity and economic performance than command and control? If intent and goal are encoded in the culture and bottom-up, not top-down it gives better context and enables us to solve tasks better. It also points to another interesting question.?

Why do founder-led companies perform better than professional-run ones??

One theory could be that with a visionary group of founders, there is less information entropy, at least for a while, and up to a certain scale, the signal is simply stronger.

With Beyond Work, we have chosen a bold mission: to learn from these systems and not just focus on software that drives more efficiency. We are designing a new kind of software that encodes intent and goals rather than ossifying structures while being adaptable and agile. Bottom-up emergent systems of work using the best from both AI and humans.?

We do this because we believe that companies that can live longer, stay truer to their mission, and be more adaptable are better not just for shareholders but also for humanity and all the people who work in them.?