Making the world a better place requires progress. Progress happens by creating technology through innovation. When people use technology to become more productive, the economy becomes more efficient and we can do more with less. The extent to which this is visible varies by industry.
The chart below shows the price changes in industries like medical care and education (getting much more expensive) versus automotive and consumer electronics (getting cheaper).
People blame bureaucracy for this difference, and it plays out when you look at the numbers. Take a look at the growth of doctors versus the growth of administrative staff:
The trends look similar when you compare the number of professors (red) and teaching staff (blue) vs the number of admin (all other colours) in Universities:
The problem with bureaucracy is that it is the opposite of productivity. Time and resources spent on bureaucracy makes the process less efficient. You need more inputs for the same output.
The world continues to innovate, but bureaucracy seems to grow at a faster rate than non-bureaucratic positions. Among other reasons, this is a failure of innovations, most importantly, computing.
Computing theoretically automates and streamlines many bureaucratic functions. For example, email is faster than mail, an online form is faster than an office queue, and internet archives are faster than physical ones.
In reality, computing hasn’t made an impact, it’s maybe made bureaucracy worse. There isn’t a solid explanation for why this is the case. As time goes on, this grows increasingly concerning. Will continued innovation in computing only lead to further bureaucracy?
Answering this question is critical for the next generation of innovation in computing: AI. Will it lead to a more bureaucratic world or a golden age of individual empowerment? Let’s explore.
Computers fail to solve bureaucracy
Computers enable people to do more with less. People can communicate across the world, share and exchange information, and coordinate better than ever. You would think this leads to less bureaucracy, that people use computers to replace bureaucracy. As shown by the numbers above, this isn’t the case.
Other people found this same lack of impact from computers on bureaucracy and productivity. Economist Robert Solow is famously quoted as saying:
You can see the computer age everywhere but in the productivity statistics.
Eli Dourado expands on this in his piece Heretical thoughts on AI:
In productivity terms, for the United States, the smartphone era has been the most economically stagnant period of the last century. In some European countries, total factor productivity1 is actually declining.
Computers existed through the growth of bureaucracy. In a lot of ways, they helped the growth. You don’t deal with people anymore, you deal with human-computer hybrid black boxes. For example, you deal with people through emails where you never know when they’re going to respond to you. Complicated applications gate information to the few bureaucrats with access to and understanding of them.
Computers haven’t made bureaucracy simpler, they make it more complicated. You often must provide more information, like environmental impact reports or security requirements, in the proper, computer-manageable forms. This makes it harder for individuals to get things done and shifts the power into the hands of bureaucracies and the organizations with the resources to deal with them.
The case against AI solving bureaucracy
AI is the next generation of computing innovation. Eli Dourado, in the same Heretical thoughts on AI piece, goes on to write about how it might not be as impactful on the big, stagnant areas like housing, transportation, energy, and health. These same areas are the ones plagued by massive amounts of bureaucracy. Paraphrasing, he says:
- Policy drives housing cost, which is set and dealt with by bureaucrats.
- Deployment drives energy costs, specifically dealing with “delay, regulatory headaches, public opposition, and lawsuits” some of which are bureaucracy.
- Transportation costs come from government opposition to changes like making cargo train sizes larger, building high-speed rail, and making supersonic flight illegal.
- Health costs come from a lack of automation, like the slow adoption of telehealth, causing an excess of people involved. Bureaucratic difficulties make it difficult to implement this automation.
None of these are miraculously solved by current AI (specifically large language models). Current AI doesn’t immediately provide a radical new way to build housing, deploy energy, create new transportation bills, or force health services to automate processes.
What current AI does provide is effective ways of dealing with bureaucracy, which might be the path to solving bureaucracy.
Outmaneuvering bureaucracy with AI
The benefit AI can provide to solving bureaucracy is simplifying the interface. Nick Arner, in LLM Powered Assistants for Complex Interfaces, writes about how people can use large language models (LLMs) to interact with complex interfaces, such as 3D modeling software. It’s not a stretch to apply the concept to bureaucracies when thinking of them as complex interfaces.
An AI can be the layer between a motivated individual or team and the bureaucracy they are interacting with. It can help automate, summarize, and translate in ways that take massive amounts of manual work. A lot of dealing with bureaucracy isn’t doing a “perfect” job, but doing a “good enough” job which AI is perfect at supporting.
In this way, AI can help individuals outmaneuver bureaucracy. Because bureaucracy is slow, a motivated individual or small team can leverage AI to “solve” bureaucracy and get things done. It provides motivated groups the leverage they need to move quickly and capitalize on opportunities. This moves power back to individuals, and away from bureaucracies and large organizations.
Examples of using AI to outmaneuver bureaucracy might include:
- Filling out paperwork such as requests for proposals, building permits, research grants (which can take a massive amount of professors’ time), and medical forms (which some claim to take up 2/3 of doctors’ time).
- Interpreting, summarizing, and asking questions of legislation, regulation, and policy which can be hundreds or thousands of pages long and few people ever read or understand entirely.
- Coordinate activism and advocacy on a larger scale with tailored messaging for citizens, bureaucrats, and officials. People get access to the context that is important to them and that helps them make better decisions.
Each of these enables individuals to spend less of their time dealing with bureaucracy. If successful, this makes them more productive. In an ideal world, this enables motivated individuals and small groups to make progress in the areas most stifled by bureaucracy like housing, transportation, energy, and healthcare.
The power of AI and LLMs is that they are available to individuals. This means they integrate with the “local knowledge2” of your work you already have. Outmaneuvering bureaucracy happens on an individual level. It requires you to take action to do it. Hopefully, this piece prompted ideas of how you can do that, and make the world just a bit better in the process.3
- Total factor productivity is the ratio of aggregate outputs (like GDP) to aggregate inputs. If improving, the processes in which inputs get turned into outputs are becoming more efficient.
- Local knowledge is what I think of as the James C. Scott concept of “metis” (from Seeing like a State). It is “a wide array of practical skills and acquired intelligence in responding to a constantly changing natural and human environment.” Also related is the Japanese word “Gemba” meaning “the actual place” (where work is done). In lean manufacturing, it means the manufacturing floor which is where the best ideas for improvement come from.
- See You Must Become a Sovereign Programmer for more thoughts on how you can make this happen.
Thanks to Alexander Hugh Sam, Nick Drage, and Charlene Wang from Foster for their feedback.
Let me know what you think on Twitter.