If we think about language fundamentally, what is it? It’s symbols which only exist in our minds.
What if we connect meaning and action to these symbols? The most prominent example of this is administration (German: Verwaltung). I don’t only mean public administration or governments, but also business administration. The kind of symbol manipulation you have to do to organize, coordinate, distribute something.
Through administration, humans can grow beyond themselves: Suddenly it’s not one person needing to speak to everyone else, but there is a central communication channel, a central register or a central instruction-provider which, based on certain previously agreed-upon rules, plus a subjective sense of obligation or desire in the followers, multiplies a command across many minds.
Unfortunately, since forever, administration has been used to centralize power. No person in power can actual hold that power without their administrators: accountants, controllers, tax advisors, lawyers.
Entire empires are built on that and they are called the “Big 4”: Ernst & Young, KPMG, PwC and Deloitte.
Right now, administrative procedures are so complex that executing them becomes very costly: Either you need to pay actual people to execute them, or there are software companies which have transformed this pattern-based procedural knowledge into code and then sell the access to this code. Other empires are built on that foundation: SAP, Intuit, DATEV and many others.
The main result of this is that large companies have a structural advantage against small businesses: They don’t need to worry about the jungle of rules and regulations because they have experts which figure those out for them. Single-person businesses are fundamentally limited in that way.
But what if they weren’t?
The primary purpose which I see for AI is not in its generative capabilities, but in its structuring capabilities. What I am already actively working on is using AI to figure out all the convoluted bureaucratic processes that prevent people from realizing their full potential. All the lack of legibility won’t go away simply by us screaming at politicians “make bureaucracy easier!”. No party will be able to do that, because the lobby of bureaucrats, of people that earn their living and have built their entire career on executing bureaucracy, is gigantic.
I don’t mean using AI as in “look Claude, do my taxes for me” and the LLM then magically assembling all the data. Instead, I mean a fundamental transition of how we currently store data overall: Away from ununified formats in disparate systems, towards each one of us having a sovereign data storage and execution environment in which applications can be run which can basically do anything.
Whether it’s “extract page 1 from this PDF” or “file my taxes”. It’s all patterns. But in contrast to how applications work right now (such as the App Store, Google Play, apt, or any individually downloaded executable), we need a fundamentally new format.
Each of these packages consists of a set of methods, and the package itself defines an ontology for all the inputs and outputs of these methods. When you install multiple applications, interoperability doesn’t arise because a higher authority defines “the one” ontology, but because applications can define “crosswalks”, mappings between ontologies. This way, your personal ontology is only composed of exactly what you need.
The methods themselves are defined in WebAssembly for maximum platform interoperability, no matter if the code is running in the browser or in any processor architecture.
Finally, everyone will have the opportunity to publish methods themselves, resulting in a global federated directory of methods. I will take it upon me to start with the basics: an ERP and everything about taxes. In the future, no one should need to pay for accounting or taxing software ever again. But of course I’m not a tax lawyer and I know barely anything about the depths of an ERP, so it’ll be upon AI to figure that out. I doubt it’ll be that hard.
Having your own business will become a fundamental human right. No one should need to be an employee.