The Crisis of Bureaucracy: Technology, Economy and the ‘Application’

The application crisis in non-profits mirrors a wider crisis of the social and political systems we have relied on for 200 years.

Hello, and welcome to the latest edition of Barely Civil Society, a window into the career-self destruction of an increasingly shrill and disappointed queer libertarian democratic socialist working in perhaps the most conservative industry in the UK. First, the monthly essay, and then some more personal nonsense is at the end.

Please like and share this post on Substack to help me break through the social media embargo that a brief refusal to enslave my content generation to an algorithm may produce.

---

The Essay: The Crisis of Bureaucracy

Introduction: The unnoticed crisis

If you want to clear a room in philanthropy or charity at the moment, say ‘AI’. As using acronyms go, it’s like saying LGBT on LinkedIn, or DEI at a White House party. But say it I must. Because whatever anyone tells you, it is having very substantial impacts on the nonprofit and philanthropy world, especially as it relates to trusts and foundations. I’m talking here more specifically about ‘LLMs’ – Large Language Models (chatbots) like Chat GPT and Gemini. But broader machine learning and ‘intelligent’ decision-making tools will become more important towards the end of the article.

But I guarantee this is much more relevant and much more interesting than you think. Because this isn’t another panic piece or a sale pitch about some charity database NOW WITH ADDED AI. This is about how it’s creating a crisis in some of the main practices we use to manage our microeconomy, and beyond that, our society. Within our industry, trusts and foundations need to understand this, even if the solutions are not simple.

But as ever, the non-profit shoptalk isn’t of primary interest to me here. (I’ll put some practical stuff on LinkedIn.) it’s about what it tells us about the wider world and what is changing for all of us. And how the culture and practices of the nonprofit world relate to our society, politics and culture.

So, yes, the kicking off point is AI in grant applications, where there has been an unprecedented avalanche of applications which has driven many funders to close their doors, repeatedly.

But the problem in the nonprofit world, just as it has been in governments the world over, is that people have been unwilling to look at the massive disruption and how to get ahead of it, rather than to watch it all slowly collapse with no plans for what comes next.

Application overload goes beyond the non-profit world and shows the coming crisis of bureaucracy itself.

The perfect storm: world context

So AI is pretty much definitely about to crash the world economy - with a ludicrous bubble which has been allowed to develop simply because it benefits the short term finances of so many ultra-rich people. Isn’t it always? Meanwhile, several industries and industrial practices have been significantly disrupted, especially where tasks related to writing and the bureaucratic process of application has taken place.

Anybody who has done any recruitment in the last year knows that the game has radically changed. AI-written applications - bad, meaningless, often heartbreaking in their naivety - have swamped the process and condemned many, especially younger, people to the scrap heap. Others are crowded out by the sheer volume. This has helped those who demand ever more complex recruitment processes where a quick letter or resume is replaced by endless tests and rounds of interviews. In fields like IT, people’s ability to use LLMs to write code, or at least to answer questions convincingly and at speed, has helped lead to live assessment, and it is common to have to endure up to ten interviews, take home tests, pair programming, and examinations.

This sits alongside a broader flooding of markets with skilled people in industries, and professions – in fact, a whole ‘middle’ and professional class of workers and a in an economy hollowing out in the centre. More and more, getting a basic job is going to be like trying to join the civil service in Imperial China.

What we’re witnessing across all these domains is something more fundamental than disruption - it’s the relationships between applicants and institutions that are breaking down. LLMs create a breakdown in trust, but through a kind of tipping over of the power relationship inherent in the systems themselves. That means much bureaucracy can no longer work as it once did, as a mechanical, rigid, system of exchange of letters, statements, evidences, memoranda and tick boxes or mouse clicks to make fundamental decisions about resources, people and power through distinct one way hierarchies. The system of exchange of submissions and decisions, or reading and writing, becomes transformed and radically undermined.

LLMs (what we colloquially call ‘AI’) are becoming the crisis point of bureaucracy itself. They form a breakdown in trust - but do this through a kind of even-ing of the power relationship inherent in the systems –That then means bureaucracy can no longer work as it once did. .

But to understand where this is leading, we need to look at specific cases.

Literary pursuits

Outside of non-profits, specific industries have been damaged by technologies that allow fast access to applications. One area I happen to know is the literary and creative magazine scene in the US. After a platform called ‘Submittable’ set up a way for writers to email hundreds of journals and magazines at once with their work, with barely any real checking, the burden was put onto the magazine. Editors returned to the age-old practice of only publishing people they knew. What was already an inwardly focused club became something more approaching a pyramid scheme. But the technology that allowed subversion of bureaucracy crashed the bureaucracy and forced its replacement with something perhaps not less fair but unfair in a different way. This changed perceptions of fairness and rightness in an industry where application or submission was a critical part. Applications and submissions, after all, are a quintessential, some might say definitive, process of bureaucracy. Submittable has since moved into bulk systems of all types of applications, including grant applications. It now markets itself as ‘Giving and granting software that never makes you compromise.’

Planning and Legal Systems

Another place we have seen this reported recently is in UK urban planning. For some time now, I have been working with people in community law who fight against city planners (those who decide what gets built or not) who quite deliberately obscure planning decisions and consultations with legal information designed to baffle and bamboozle. This benefits those with massive legal teams (such as massive property developers), but disadvantages residents in areas destined to be gentrified and ethnically/socially cleansed.

My own discovery was that, as I undertook a research project on community activism for a law centre, with the help of the UK Law AI bot I was able to understand and translate legal information (confirmed by the legal experts I was working with) in ways I could understand. This could, I have since advised campaigners, help them to at least understand systems designed to exclude them. They could even use it, perhaps to generate letters in the baroque formats required.

Recently in the Guardian, a typically Millennial headline claimed that AI systems now offering translation and submission of objections as a service for a fee (as usual, some techbro dickhead comes along and charges you extra money to do what normal LLMS will do out of the box) were simply designed to allow mass ‘NIMBYism.’

Well, not all planning objections are from NIMBYs, Guardian, and you would be the first to complain about many of the shocking environmental and human rights abuses by developers that activists have to contest. But also, this practice offers a kind of even-ing of the odds, a rebalancing of the power, and a sudden ability to circumvent some of the guardrails of the bureaucratic juridicolegal process. That may indeed allow the gumming up of an already creaking system. It evens the odds, and rebalances power, but the point is that a power, and resource imbalance is all that ever held the social practices of bureaucracy together. When power is pushed to crisis, however, it has to make decisions about how it will respond. Because it is powerful, that often goes the wrong way for those who came to challenge it: for example, shutting down or circumventing any remaining avenues for challenge through the system they have lost control of.

The non-profit world

And now this same pattern is hitting the charity sector. The ‘Oh shit’ moment á la Brexit has caught them unawares. The closures of funds due to up to 400% increases in applications are indeed heavily influenced by this.

Let’s be clear, the massive crash, panicked-applications, and tsunami of refused requests is not just a result of AI, any more than the macroeconomic tanking economy and jobs market is. Worldwide economic collapse, underinvestment, and reduced giving, all play their part in the contraction happening at pace in the non-profit world, including Trump’s dismantling of the USAID programme which has affected non-profits globally. But the point I’m trying to make is that the wider and industry-specific economic and cultural conditions have combined with a technological change to create a perfect storm. A few foundations who now who say they are ‘not seeing it’ are, precisely, just not seeing it - that’s not the same as it not being there. And the whole point of LLMs is to be convincing.

What will happen? Hard to say. But I already see some changes which I think have much wider significance for the whole of our culture. One is a potential return to individual relationships, and ‘who you know’. Already I am seeing this in funders; most of the recent successes I am seeing are with organisations winning money from those they already know. In many ways, this is what people have been asking for, for years, when they ask funders to fund longer and larger, and with fewer applications….

It’s a return to patronage. But to understand why such a return would be so significant, we need to understand what bureaucracy was meant to solve in the first place.

What bureaucracy was meant to solve

Modern bureaucracy developed, but especially took off, in the West alongside liberal democracy. The idea was that managerial approaches could allow access to juridicolegal systems that would previously have been available only to those who had personal contacts and connections with kings. By filling in a form, you were able to join a system that, on the surface, made decisions without prejudice. The sheer formality and managerial remove meant that a more equitable system - and one which privileged the literate middle classes - was available to anyone who could write.

Of course, part of that system was always a kind of epistemological coercion that prevented or delayed access to favour, or indeed justice, for those least well-equipped to deal with bureaucratic systems. Those people included the poor with oral cultures, colonised peoples with a different language, working class people without a high level of literacy education. Bureaucracy was both an expansion of access to fair process and an attempt to contain it. Remember that another word for, say, an application is ‘submission’. The word itself betrays its origins in what it is meant to replace, but also to continue by indirect means. (That is, direct supplication to a lord or king.)

Practical implications

So what’s happening in the philanthropy sector’s bureaucracy? In the microcosmic world of the day-to-day charitable philanthrosphere, these changes are afoot.

Funders are starting to see AI-generated bids. Some are poorly written - the absolute guff I’ve seen is similar to what I’ve seen in recruitment. But bad applications existed long before AI. AI can be used to fabricate or exaggerate claims. But people could always do that: it’s the number one risk of fundraising and funding. But there’s a bigger thing at play here, as application systems are being overwhelmed by mass input. And this brings us to the most likely outcome for our whole technological society.

The decentring of the sector means key policies around LLMs will be decided piecemeal. The National Lottery Community Fund is now accepting bids by AI, but not by actual humans – if they are those pesky parasites of the nonprofit world, ‘consultants’ and freelancers, who now make up a massive proportion of the workforce. More broadly, there has been little to no leadership from the big beasts of infrastructure (who are now not so big). People like me and the Institute of Fundraising have tried to get bid-writers to use LLMs ‘responsibly’. For me, using LLMs properly means doing pretty much the same work as I have all along, but more efficiently, and sometimes in a more ‘collaborative’ way with the LLM as an editing partner.

Increasing efficiency in applications is what we have been demanding funders allow us to do for years.

And indeed, funders have been claiming for years that they wanted that too, especially for those without professional bid writing skills. But what replaces the application ‘process’ may be even less fair than knocking on a door and handing in your 20-page letter to Santa with full costings. Be careful what you wish for.

As new systems take hold like two-stage processes, and many funders start to expand resources by taking on additional outsourced bid assessors, there remains a more fundamental problem: that in a system where bureaucracy’s power dynamic is evened out, the systems no longer work. This could lead to collapse – quiet in some cases, or ‘loud’ in others.

Where does this end?

One of the goals of bureaucracy was also to remove the ‘subjectivity’ and ‘emotion’ in decision making. This always has been one of the fundamental dreams of bureaucracy - a bloodless, fully ‘objective’ reason above favour or personal sentiment. Of course this brings significant risk and dangers of its own. That modern bureaucracy developed alongside science, with its focus on rationality and process, is no accident; nor indeed is the later development of ‘management science.’

As we move now to become the ‘Impact Industry’ (the latest term being used to describe people who try to make the world a bit better in an institutional setting), this desire to neutralise decision making will only be stronger. Indeed, ‘social investment’ is bound to increase the sheer quantification and dehumanisation of processes as much as ‘performance’ (how many, and how much?).

The thing is, most human beings don’t like making decisions, because they then have to take responsibility, and, god forbid, be accountable for them. The more we hide the decisions, the more we can set the terms of the system and claim bad decisions are accidents. Or indeed, that they are unpleasant for some, but the ‘right’ ones ‘objectively’ - as decided by an algorithm. The implacable ‘black box’ of bureaucracy – the Kafkaesque citadel of filing cabinets which can only be accessed by a phoneline or a form sent up an airtube – potentially comes of age with AI.

‘Faceless’ bureaucracy is meant to make sure nobody needs to see the applicant cry. A pen stroke, key strike, and an email that means Number Go Up Or Down – this is a profoundly comforting things for humans in power. It’s how people who aren’t complete sociopaths manage to do pretty awful things. And the question of to what extent we want to remove the ‘human’, emotional, or subjective element of decisions that bureaucracy promises, has been something of an ongoing theme for humankind for 200 years or more.

In the grant-writing/ awarding sphere, deciding what to fund is less often, although not never, such a matter of life and death or overt ‘violence’ as, say, the use of bureacrcy in wartime. But it certainly does have very live and real effects on individuals and communities. Where bureaucracy originally seemed to promise fairness by removing subjectivity, we might also ask to what extent, or to what level, the removal of subjectivity – the actual thinking of a human agent - is actually desirable.

Solutions? Or more problems?

In the non-profit foundation world, we’re heading towards a situation where writing bots will meet reading bots. We will start to see AI being used to read and decide on bids as well as write them. I know for a fact that is already happening on an individual basis as people use LLMs to summarise applications they receive. (An irony - long bids are demanded by funders, which are then too long to read....) That process then writes all manner of truncations, themselves with selections made by AI, which will undoubtedly affect the final decisions. But I think it is highly unlikely that is where it will end, and we are facing a situation where at least initial ‘sifts’ are done automatically by AI and machine learning. This has long been the case in recruitment, especially in the US.

Much as we developed systems in marketing to craft SEO junk to be crawled by bots, which no human will ever read, this creates a fundamental question: do we need the process of reading and writing at all in ‘application’? If no human is involved, is application and reception the best method? Does bureaucracy need a human? Do we just let the black box do its thing and achieve the ultimate removal of subjectivity and embracing of an ‘objectivity’ which can in the end only be embodied by a machine? (Leaving aside the very obvious fact that these machines are far from without bias.)

If so, what does that look like? It is likely to be some kind of passive crawling of databases. Some tech people already have this in their sights, including, it would appear, Submittable, as well as at least one UK database tool. Their plan is likely to create a system that becomes the absolute interrelation and interlocutor between funder and fundee, applicant and institution, job-seeker and employer, from invitation to to outcome and beyond. Make no mistake, their goal is to replace all fundraising or recruitment – with themselves taking the money ‘saved’. The fact that these companies sell a perfect system which allows more room for ‘relationships’ sounds great. But exactly how that works is a different matter.

The Crisis of Bureaucracy?

So here we are at the fundamental point: LLMs have created what I call the crisis of bureaucracy: the old administrative systems (forms, long applications, human review) are being overwhelmed because everyone can now generate text, and make ‘submissions’ or ‘applications’ at scale.

And for me, the greatest disruption that LLMs have produced so far is not some kind of major cost saving or transformation of working practices in organisations. (Indeed, research recently showed that there was little saving within organisations – tellingly, because the effort of writing was removed, and transferred to the reader, who had to read absolute bollocks created by an LLM.) It’s in the relationships between ‘applicants’ and ‘institutions.’ The mismatch of resource: a substantial, if painfully slow machinery for reading, set against the effort premium of individual writing, was essential to the power balance that bureaucracy and the power regimes it held up relied on. The gatekeepers have lost their gate. Time to build a new one.

The point is that while we are seeing massive disruptions for funders in the UK (with more to come), the crisis is not just in funding applications. It is in the very idea of ‘applications’ across our society, right at the heart of bureaucracy. Jobs, legal discourse, academic pursuits, banking and investment of course, creative arts, goodness knows what else. What we face is logjam, power upturned, and the high likelihood of the more powerful gatekeepers shifting strategy entirely.

So what is this strategy?

In all of this, I see two possibilities. This may be the final apotheosis of bureaucracy. The bloodless removal of human subjectivity, with everything decided by an algorithm.

Or it could be the last crash and crisis of bureaucracy, where bureaucracy as a system can no longer cope with volume and the even-ing out of systems power between applicant and institution. That leads to a world where bureaucracy evaporates, taking us back to… what?

Well, what it replaced. A system of relationships and patronage. Grace and favour. I don’t think any of us could wish for that the process involved fewer human relationships in most aspects of our work. Indeed, grant-makers and even recruiters talk more and more about the importance of the ‘relational’ approach. But what kind of human relationships will it take us back to? And what power relations will that create?

So patronage, ‘who you know’ and the sheer hustle may become even more vital. And indeed, at the front of the queue will be those who learn this best, or who are put there by power imbalances at the outset. Class, race, gender, sexuality, all of these things will be affected. I don’t suggest that these relationships cannot be managed or that we cannot seek to introduce a certain ethical balance and some level of ‘fairness’ to any return to subjectivity. (And we will no doubt claim this process is ‘objective somehow – it’s one of those words and concepts we can never reach, and yet remain relentlessly focused on attaining.) But it may be that the human relationships it takes us back to are far from what most of us, even those of us who innately distrust bureaucracy, would want.

The fundamental questions at this point: is this the crisis of bureaucracy, or its apotheosis? If no human is involved, is application and reception even a meaningful process? And if not, what replaces it? And does this restore humanity and save effort, or entrench unfairness and exclusion from power?

---

The (not very) funny personal bit

Last week I chaired a session on radical ideas for trustees at the Festival of Trusteeship, where several people presnted such radical ideas as, you know, maybe trustees doing some training every now and then. You may have read my previous work questioning what role trustees ought to play - and have played, over the least thousand years. I did end by suggesting that perhaps the whole shitshow was too shitty to continue the show, but subtly. You can see it here - some great ideas to be tried while we rearrange the deckchairs, anyway…

Some great ideas.

Then they let me off the leash.

Shout out to Hannah Kowzun, who won the vote and tore standard trusteeship a new one. We may have to set up a nursing home for exhausted and angry nonprofit people. Let me know if you can fund it - maybe the Mercers, if you’re reading?

Prir to this, I spent two weeks on holiday. A few days in Kent (a weird graveyard), and then 10 days in London on a staycation. Mostly,I watched horror movies and read Hegel. I also visited several cultural institutions and now have a second mortgage to pay off. The Transport Museum refused to give me a local’s discount, because apparently that is only for people in North London. ‘Oh,’ I said - ‘you mean like the tube?’ They didn’t laugh, and I was in a foul mood, but improved it by glaring at parents who let their children run around and enjoy themselves. How dare they.

Meanwhile in politics, Starmer is in trouble. Sadly, the two potential frontrunners to replace the stiff-with-a-quiff we currently have are even further to the right. Streeting is a neoliberal worse-than Blairite shill, who just enjoys scorched earth politics, and Shabana Mahmood is another entrenched social conservative.

I’m hoping Keir stays until Burnham can get back into parliament. But can you imagine them letting him back?

Meanwhile, has anyone else noticed this about Wes Streeting (see image)? Amusingly, he congratulated Mamdani online and said he felt hopeful to see new voices breaking through against the forces of conservatism. As somebody noted on social media, he seems to have failed to notice that, in this scenario, he is Andrew Cuomo.

Wes, and Wes

And finally - watch to the end….

@thelondoneconomic

The London Economic on Instagram: "Watch until the end 👀"

See you in two or three weeks.

- Alex

---

Barely Civil Society is about making the nonprofit world braver.

Here are three ways to help

1. Subscribe!

Subscribe and share! To receive new posts and support this work, consider becoming a free or paid subscriber. Or share this post.

2. Share!

I’ve taken a four week break. Substack’s algorithm will crucify me. Please share to give me a chance of anyone seeing this.

3. Buy me a beer!

Bugger coffee. In this climate, you need buy me a beer. That way you can help me spend more time researching and writing - including some bigger projects I’m working towards. And keep me drunk.

They’re a bit tiring on my neck, but I do feel so masterful when I wear my antlers.