The Death of the Corporation
"A corporation is an organisation—usually a group of people or a company—authorised by the state to act as a single entity ... and recognised as such in law for certain purposes"–Wikipedia
Corporate survival rates aren't what they used to be. A quick look at corporate lifespans for the S&P 500 shows a downward trend over the last 60 years:
It got me thinking–what will be the impact of advances in AI on the textbook benefits of the large-scale enterprise*? Are economies of scale and access to the best talent still key, or are we in the middle of a massive paradigm shift?
In other words, with Large Language Models (LLMs) redefining the economics of human capital, what does this mean for the knowledge workers that make up the bulk of the large-scale enterprise workforce?
Let's look at a couple of statistics before we try to answer this question.
| Small or medium sized (SME) | Large scale enterprises (LSE) | |
|---|---|---|
| Employment | 69% | 31% |
| Turnover | 59% | 41% |
| Exports | 45% | 55% |
2020 averages for all 26 OECD economies (source: OECD Library).
At the national accounts level, LSEs generate more turnover per employee than SMEs and have a far bigger share of national exports.
The modern corporation has stayed true to its roots as an international player.
*) I'm using the OECD definition: 1-249 employees is a small or medium-sized enterprise (SME) and >250 employees a large scale enterprise (LSE).
The New Truths
On the other hand, corporations also operate in a rapidly changing environment:
| Competitive Advantage | The Old Firm | The New Truths |
|---|---|---|
| Manufacturing and Production | Economies of scale | Supply-chain management |
| Marketing and Advertising | Consumer brand | Personal brand |
| Communication and rel. mgmt | Social network | Artificial intelligence |
| Business Management and Innovation | Knowledge factory | Algorithm factory |
| Information Technology | Software | Infrastructure & hardware |
| Market position | Physical product | Service, digital-first |
| Valuation driver | Combined assets | Intellectual property |
| As an employer | Stability | Status |
The core organisational principles of corporations are changing over time.
To point out the obvious, in ye olden days corporations relied on people–lots of people–to remain competitive. In contrast, the power-brokers of today have built their market positions on a completely different set of competitive advantages.
As a result, we no longer have small-country-sized behemoths like General Electric that are able to dominate markets by sheer force of numbers (human and capital).
Even though I couldn't find data on this, I'm willing to bet that the average revenue per employee for the corporations listed in the S&P 500 has gone up a lot over the last 50 years.
The biggest companies in the world need less and less workers.
They are able to leverage software, IP and infrastructure plays with great success. And it looks like AI will give them an exponentially bigger lever.
In this changing environment, the safest bet is the infrastructure play–as can be seen by the market valuations of the "Magnificent 7".
And infrastructure plays naturally tend towards monopolies.
Get That Monopoly Money
The most common justification you'll hear for these monopolies is that LSEs need to extract rent in order to be able to invest in innovation and to drive progress.
This debate is also happening in the AI community right now, where closed-source and open-source proponents of AI and AGI both have valid points.
I personally have always been a big open-source advocate and user, because I am a strong believer in open innovation and common progress.
From a capitalist perspective closed-source also makes little sense. The biggest market by far is consumer households.**
So whatever you build or sell, if it doesn't directly or indirectly cater to the biggest economic force in the world, it will have a much smaller chance of success.
And with millions or billions of users, there is very little in terms of secrecy that you can do as a company. I haven't seen a product or service that couldn't be easily reverse-engineered once it's out there.***)
There is a common jibe making rounds on social media that nobody really knows what SalesForce does. The same holds for Oracle, IBM, HP and a ton of others.
From personal experience, the 80/20 rule holds in most corporations. 20% of the people are responsible for 80% of the value creation.
As an LSE if you don't have an infrastructure or marketplace play it is going to be much harder to remain competitive. Profit margins will decrease, brand value will suffer, and employees will become less and less enamoured with your company.
The only way is to leverage your scale is through acquisitions–a strategy that Microsoft has mastered skilfully in the last decade under Satya Nadella.
One of the smartest thing of his acquisition strategy in my opinion is that MSFT has kept Github, LinkedIn and ActivisionBlizzard distinct brands and companies, rather than folding them into the corporate umbrella.
This has been a huge part in the success of Microsoft, along with their AI strategy.
But they have been the exception–a lot of other LSEs have hit rougher weather.
They are struggling to make old plays work.
And I don't think this will change.
From personal experience, the 80/20 rule holds in most LSEs. 20% of the people are responsible for 80% of the value creation.
The rest is on board for the "corporate welfare train".
And at some point, that train will run off the rails–as can be seen from the decreasing corporate lifespans in the S&P 500.
What it comes down to is that bigger is no longer always better.
Another sign is the decreasing number of IPOs and lower IPO proceeds in recent years–I haven't done a proper analysis here, so the following is purely conjecture: it looks like the market is pricing out size and scale as valuation multiples, which lines up with the reduced profitability for companies of doing an IPO.
What it comes down to is that bigger is no longer always better. In fact, it is often worse because of the overhead and complexities involved in managing these kind of massive organisations.
**) Although that might change at some point, if we ever get to something like "superintelligence". I've done a Youtube video on this some time ago.
***) Which doesn't mean I don't believe in the protection of intellectual property–only that any truly ground-breaking innovation needs to be adopted for it to become worth anything. Cf "The Diffusion of Innovations" by E. Rogers.
The dissolution will happen naturally
What I think will end up happening is that most big corporations will turn out to be a bunch of smaller companies.
And that organising R&D collaborations in networks, supply chain with software, and marketing with AI agents is a lot more efficient and will yield better results.
That you'll see the rise of boutique R&D firms catering to multiple companies. Companies that get their competitive advantage from their go-to market implementation and customer service orientation.
This is already happening in AI and healthcare, arguably the two most important industries of the 21st century.
Boutique research labs develop deep tech solutions that they monetise through both direct-to-customer services and licensing contracts with other companies.
What I think will happen is that most big corporations will turn out to be a bunch of small companies.
To be clear, this is a slow-moving trend. Even for AI, we are talking decades.
Back in 2014 we had pretty decent computer vision algorithms, but skip-jump to 2024 and we still don't have fully autonomous self-driving vehicles.
And there will still be a couple of massive winners. Companies that recognise they need to rewrite the rules of the game. Companies that, to stay with the analogy, build the railroad tracks for autonomous vehicles.
But outside of infrastructure plays, what we'll end up with is... just companies.
Killer marketing departments spun off to cater to commercial clients. A great supply chain department that hires software developers and becomes a SaaS company. The best finance departments become boutique consultancies. And so on.
Rather than keeping internal departments afloat on corporate profits even though they are providing mediocre services, I think there will be a trend to "outsource" more and more business functions as the quality of commercial offerings goes up.
In my mind, a strong driver in the cost / performance tradeoff of these commercial offerings is the rise of AI agents.
Agentic AI applications are often too expensive to build internally. Yet they have the potential to outperform 80% of the corporate workforce in the not too distant future.
Agentic outsourcing could be the catalyst for a larger trend in which corporations need less and less people to run their operations.
This could accelerate the shrinking of corporations up to the point where they are no longer corporations, but just plain old companies.
On a lighter note, a potential upside of these developments is that there will be much less need for middle management.
LLMs for middle management 🇧🇧
Because even if it sometimes feels like management has been around forever, it is actually a fairly recent innovation. In 1985, Peter Drucker mused that
"[m]anagement is the new technology ... making the American economy into an entrepreneurial economy. It is also about to make America into an entrepreneurial society. ... [which] requires above all application of the basic concept, the basic techné, of management to new problems and new opportunities."–Peter Drucker (1985), Innovation and entrepreneurship
Unfortunately, this hasn't really panned out. In my experience, most LSEs have a management style that is closer to feudal fiefdoms than entrepreneurial brilliance.
Somehow somewhere, adding additional layers of bureaucracy hasn't been the modern-day equivalent of sounding the golden horn.
You only need to look at medieval Europe or feudal Japan to get a sense of the ramifications of a conquer-and-divide strategy for the power dynamics in an organisation.
So maybe it's not such a bad idea to start using LLMs as drop-in replacements for middle management. They can also talk forever, don't create measurable business value, and occasionally say really smart things.
Sources
Kevin Beasley
Department for Business, Energy & Industrial Strategy
Kelly Main
Learn more
Ollie Forsyth
George Chan
Member discussion