It starts with pre-filling your forms
And it quickly becomes Fully Automated Luxury Communism
One of the things you discover when working at a large corporation is that promotions are not just something that happens; there’s a process, and that process is all consuming several times per year.
It‘s all about feedback
Most large companies do half-yearly (or if they hate themselves, quarterly) “review cycles” in which managers write up performance reviews of their staff, based on the feedback the staff member has given, the person’s meeting of their objectives, and so on. Managers generally need to decide if the staff member has met or exceeded the expectations for their job and level, as set out in their job’s overall profile. Each kind of job has certain performance indicators, and feedback generally gets classified against which indicator they align with or oppose.
Based on all of this data, the manager writes up a recommendation as to whether the employee warrants promotion, performance management, or something in between. These recommendations get shunted up the stack to managers who need to make promotion and bonus decisions based on the feedback from the managers who report to them. Not only do they need to make those decisions, but they often need to defend those decisions when budgets for bonuses and promotions are constrained, as they always are. They defend those decisions based on the recommendations that have been provided to them.
Managers hate admin as much as the rest of us
Managers are people too. They hate filling in these forms just as much as anyone else does. The whole experience is wretched, and is especially wretched for managers, in my opinion as someone who gave up being a manager years ago.
Across the corporate world, there’s been a deluge of “Ooh AI, what can we do with this weird new tool?” funding, pretty much all of which can be summed up as smarter auto-complete for form fields.
All form fields that can be populated by AI, will be populated by AI
Every field that used to need someone to fill in will have a little ‘complete this field with AI’ button of some kind. Maybe it will have a little contextual menu letting you fill it in in different ways. Behind it will be a basic RAG API talking to an LLM using relevant data as system context.
There’ll be fine-tuned prompt language aiming to reduce bias and ensure diversity. Each field will be auto-completable with the most perfect corporate-speak, something LLMs do really well. All the manager will have to do is read it and tab to the next field. At the end of the form, the manager can just scroll past some boilerplate about needing to review any and all AI-generated summaries and click submit to approve them all (i.e. take responsibility for them).
This is already happening at large firms like Accenture. (I used to work at Accenture.)
What’s a knowledge company?
A typical knowledge company sells access to structured knowledge in the form of software, processes, designs, and presentations.
A software company, design house, or consulting firm will have engineers, analysts, designers, writers and creatives, lawyers, and other specialists who form the firm’s beating heart, its “means of production”. Scattered through the body corporate, you’ve got managers and admin staff, the connective tissue, life support, and waste management without whom nothing would get done, and also floating around, you have the marketing and vision people who steer the ship at a higher level, in tune with customer needs.
Companies exist within a network of customers, shareholders, suppliers, competitors, threats, and opportunities. If a firm is sufficiently large, then there is likely a lot of crossover between customers, shareholders, suppliers, and competitors.
So money comes in from customers, rains down on the staff, who absorb most of it, and the plant and equipment which absorb some of it, and the rest heads out the door in external expenses, tax, and returns to shareholders.
AIs have already shown themselves to be okay at basic creative tasks like writing, copy-editing, software development and debugging, contract summation and preparation, UI development, and many more. As AIs get more capable, and the range of forms that can be autocompleted gets more and more intricate, it doesn’t take long and a lot of jobs go the way of the typing pool. With fewer jobs directly wired to the firm’s products and services, there’s less need for all those connective tissue roles either, and it’s not long before, instead of auto-completing forms, those forms just go away.
Full automation
Extrapolate increasingly sophisticated AIs, and the end product is surely a firm with no staff, just an AI dedicated to deciphering incoming feedback, then generating responses and improvement requests which get sent to an AI dedicated to coding up changes to the firm’s products and services.
Customers pay via smart contract, and funds are disbursed in the same way, authorised by the feedback AI.
In this model, money comes and goes via a smart contract (likely mediated via an app, API, or via another smart contract), while feedback comes and goes via an API to a bot, which has functions it can invoke to instruct the code updating AI.
At scale, across whole knowledge industries, firms whose products are services that enable workplace functions, like CRM, HR, training, and recruitment, or catering, will become increasingly boutique; then those boutique firms will consolidate; and then most of them will vanish like a bromide camera looking down the barrel at ink-jet printers.
The majority of a firm’s operating expenditure, previously spent on staff, can now be spent on running the AIs that are improving the services offered to customers, competing with competitors, nickel-and-diming suppliers, minimising tax, and returning value to shareholders, most of whom in many cases used to be the firm’s employees.
It doesn’t stop with knowledge work
Advances in 3d printing and other advanced manufacturing techniques, as well as drones, energy generation, storage, and supply, as well as transport and robotics, are happening at a similar pace to advances in AIs that write poetry to your girlfriend and explain the Higgs field to your niece. In fact, these worlds are coming together with advanced cognition now being embeddable in mobile devices. Companies that require the mobilisation of human and animal brawn will increasingly supply that brawn via robots and molecular manufacturing in the service of their customers and shareholders, following instructions from corporate AIs.
Company directors have a fiduciary duty to maximise value for their shareholders. A few also have embedded broader societal outcomes into their corporate charters, but by and large, all corporations are required to maximise value for their shareholders. This will ultimately necessitate board members voting themselves off the board, then dissolving the board and putting decision-making power into the hands of the company’s AI. Legally, they’ll have no alternative, but I expect they’ll be remunerated handsomely to make up for it.
Once a company successfully heads down this path, unshackled from human constraints, the only thing preventing a rush on entire industries will be government regulation and national protectionism as international safeguards are rushed out. A multinational race to dominate industries will emerge out of seemingly nowhere, until probably only five corporations are left standing, and none of them employ anyone. This will work out about as well as you’d expect.
The only actual employers remaining will be government, judiciary, and law enforcement services; and only then because of tradition. Most people will live lives with the lower layers of Maslow’s Needs pyramid accounted for by the mostly automated state.
Fully Automated Luxury Communism
The next stop is Fully Automated Luxury Communism. But the road there is going to be messy. Assuming we survive the transition, and against a rising backdrop of global warming, climate instability and unpredictability, and resource depletion, as well as tensions between nuclear powers, I like to think there’s a post-scarcity future worth striving for.
People who work for firms that survive the transition, and that have generous employee stock programmes, will be considerably better off than those who work for firms that do not. Similarly, investors with shares in firms that make goods and services that people will actually pay for should do well.
Machines and personal AIs will increasingly help people navigate this transition, while quantum leaps like high-fidelity protein printers, house printers, and humanoid robots that can hang your paintings, get a ball off the roof, and keep your house clean work to upend society while we watch.
Most economists theorise that some sort of universal basic income will help, paid for by taxes on automated wealth generation, and maybe that’s true. Certainly something needs to be done, and fairly fast, as the genie is out of the bottle now, the train is hurtling through the tunnel, and for many, the light at the other end is probably another train.
Links
- https://newsroom.accenture.com/news/2024/accenture-report-finds-perception-gap-between-workers-and-c-suite-around-work-and-generative-ai
- https://newsletter.pragmaticengineer.com/p/ai-coding-agents
- https://en.wikipedia.org/wiki/Fully_Automated_Luxury_Communism
—
Updated: 2024–06–05 to include a link to https://newsletter.pragmaticengineer.com/p/ai-coding-agents
Updated: 2026-02-14 to fix some wording and grammar, and unlock the paywall since no-one was reading it and I’m terribly bad at promoting my own writing.
—
Like this but not a subscriber? You can support the author by joining via davesag.medium.com.
