Fireflies and algorithms — the coming explosion of companies
Company formation has, so far, been relatively unaltered by the data revolution.
But change is coming, and will likely create a massive increase in the speed, number and complexity of companies, all powered by automation of company formation, and driven by demand from new technologies.
As with the shift from paper to electronic stock trading, the automation of companies will have profound and wide reaching effects for society as a whole, fundamentally disrupting how, when and why companies are formed… and the ways we hold them to account.
This blog is the second in a series we’ll be publishing about OpenCorporates Phase Two: the transformation of OpenCorporates in a rapidly changing corporate world. Read the other blogs about Phase Two here.
For something that is entirely virtual, company formation is remarkably stuck in the world of paper and humans.
Companies are nothing more than legal constructs. They are artificial persons created by states or courts, and given the power to enter into contracts, own assets, take on debt. You can’t see or touch them — though you can see the things they make, or the offices they own or rent — they only exist in the legal world.
Despite this, the way companies are created and dissolved, and the way we collect official records about them is still largely unchanged from 20 years ago. In most countries, even those that are heavily digitised, creating a company is an almost entirely manual process:
- A person wants to create a company.
- That person (or a representative of them) sits down and fills in forms, either on a computer screen or on paper.
- They send them to the company register or court, where they are examined and processed, again largely manually.
Compare this process with, for example, that of listening to music on your phone. You want to listen to something new: you might have a song playing in the background and fed it to Shazam; you may have seen one in your social network feed; an algorithm may have analysed your previous listening choices and identified something it thinks you’d like. Within a few seconds, you’re listening to a new track simply by clicking on a button. It just works, and you hear a song you love.
What’s happening in the background, of course, is a hugely complex system, involving tens, maybe hundreds of different companies and contracts. These entities are connected via a web of APIs, sending each other data underpinned by digital identities, authentication processes and strong encryption.
If you ever stop to think about all that data, some of it about you, all those agreements, the opacity of it all, you may experience a moment of unease — but it usually quickly passes.
Incorporating a company is rather different from listening to a song. Companies have legal personality and are ‘independent’ actors. They can create value, they can make great products, they can employ people. They can also fire people, can cause financial crises, can launder money, commit crimes, or even kill people.
Nevertheless, just like an audio file, companies are virtual. So it is curious that the process of creating them, of dissolving them, of submitting statutory filings has so far been remarkably unaffected by the data revolution.
A concrete example:
Over the past few months, OpenCorporates has created a number of new companies as part of the restructure to safeguard our important public-benefit mission for the long term.
It was a remarkably manual process.
A piece of paper was sent through the post with a written signature that could easily have been forged¹ – there was no authentication, no use of digital identities or digital signatures to confirm identity. There was no opportunity for the director to be linked to directorships of other companies, which would have improved the information available to users, and also alerted the director to potential identity theft (indeed there have been plenty of examples where the processes allow directors to be falsely appointed to companies, or where fictitious addresses have been used, or where control of companies has been passed to criminals by these lax processes).
To be fair to Companies House, the UK company register, it was sent by paper by OpenCorporates’ lawyers, in part because the Articles of Association and other documents were non-standard. However, even if the filing had been submitted online the same issues would have been there.
This situation is crazy in our data-driven world. It is also unsustainable, not because it’s irrational – although it is – but because there are powerful forces that will change this, and in the process change the very nature of companies and company information. This change is coming, like it or not, and the result will have profound implications for society, not all of them positive by any means.
The coming company revolution: demand
We have identified 4 key drivers that will create a strong demand (the ‘pull’ factors) for automated legal entity systems:
1. The need for unambiguous signed data.
As we’ve seen, company incorporation is almost completely a manual, trust-based system, and in some jurisdictions an opaque one too. Not only does this have many negative consequences, allowing criminals and corrupt politicians to use companies for illicit purposes, it also makes the data unsuitable for new blockchain-type technologies, which require clear, unambiguous, provenanced data, backed by digital identities.
This emerging need is one of the many reasons OpenCorporates takes provenance, understanding, and correctly mapping company data so seriously. Of course, the primary reason is because it’s the right thing to do.
There are also some in the law enforcement and anti-corruption communities who are investigating how such technologies can fight criminal activity related to companies. It’s harder (though not impossible) to pretend you’re Donald Duck if you have to tie that to your official identities.
2. The need for legal entities on the fly.
Smart contracts (contracts written and executed in computer code) have had much less hype about them than Bitcoin/Blockchain, but the potential to be even more disruptive. Turning many major (and minor) transactions between two or more parties — with all the potential for fraud or disagreements — into code-supervised events that are completed automatically will have profound consequences.
In fact, companies are already incorporated for specific deals in the financial markets — to limit risk or to gain financial benefit. Currently, due to the cost and friction involved in doing so (getting high-paid lawyers to do it), this only happens for relatively large deals. However, put all this into say, a Software as a Service (SaaS) system, and that cost would come down by 100 or even 1000 times. Companies will be created as needed, for insignificant costs, and without the friction of human intervention.
3. The need to improve process.
As the world becomes data (goodbye letters, goodbye cash, goodbye manufacturing jobs, taxi/truck drivers, banks…) the parts that are still analogue will increasingly be undermined as they become a barrier to efficiency and automation. Having a high-priced lawyer sit down and fill in a company incorporation document – taking a day or so to do so, when the rest of the process takes seconds and costs a few pounds/dollars/euros – will be too great a barrier to be sustained.
4. The need to ring-fence liability issues introduced by new technologies.
Limited liability companies are, as their name implies, designed to limit liability to the assets of the entity, leaving the owners and managers unscathed. It’s a curious concept, to say the least, but when not abused, and when not surrounded by opacity (which provides fertile ground for bad actors), they can be an extraordinary vehicle for innovation and prosperity.
Originally the need for limited liability companies was to build a railroad, or a factory; entrepreneurs needed to raise money to build them and investors could limit their losses to what they had invested. Now limited liability companies are used for everything from corner shops to special purpose vehicles.
But as we move into a driverless future, will limited liability companies be used for self-driving cars, which are, potentially, liability minefields? If major merchant ships already have a unique corporation associated with them, why not self-driving cars, particularly given they may well be subscribed to rather than purchased – the user will not be an owner with a title deed, but a subscriber with an End User Licence Agreement, just as with software.
If self-driving cars, why not personal body sensors, or smart limbs, particularly if the corporation can be cheaply created on the fly by software?
Specific laws may be introduced to address the issues and provide certainty in the field of self-driving cars, but the Internet of Things, particularly the AI-powered IoT brings a host of similar issues. Decisions will be — and are being — made by code which interacts in complex ways with equipment’s hardware and the surroundings, and has significant real-world consequences.
The coming company revolution: supply
In addition to the pull factors (demand), there are a number of push (supply) factors that will address the problems experienced above, including:
1. Blockchain/DLT systems.
A number of company registers are exploring Distributed Ledger Technology. While we think that the benefits of this technology are somewhat overhyped, some of the underlying technologies, including digital identities/signatures, immutability and clarity are relevant to the incorporation process.
For an example of this, see the Delaware Blockchain initiative — although interestingly Delaware seems to be backing away from it somewhat, after they realised it might damage their ‘business model’.
2. Competition between jurisdictions.
One of the challenges for society in this global, corporate world is the ability for jurisdictions to compete with each other for a slice of the incorporation business, creating a race to the bottom. This competition has hitherto been primarily around secrecy — and to a lesser extent tax and regulation.
Jurisdictions such as the British Virgin Islands and Caymans compete both with each other, and with onshore secrecy jurisdictions such as Delaware, to tempt companies and people to incorporate entities there. A vast amount of money flows through secrecy jurisdictions, but only a tiny proportion of this goes to the country itself. For example, company registration brings into the Cayman Islands only $125m in order to get the crumbs brought by the incorporation business. For more on this, see Nick Shaxson’s work on the Finance Curse.
We believe this competition will start to get interesting — and cause a shake-out in the market — once technology becomes a factor, perhaps with several jurisdictions collaborating to allow on-the-fly corporate networks to be created.
3. Lobbying from interested parties
Those trying to evade, for example, the requirement for Overseas Territories to publish public registers of beneficial ownership (i.e. who really controls and benefits from companies) will lobby to create loopholes which technology will fill.
One example could be what we’re calling ‘firefly companies’ — the blink-and-you-miss-it scenario brought about by ultra-short-life companies, combined with registers that remove records once a company has been dissolved, meaning that effectively they are invisible.
4. Technology to get around the problems and bottlenecks
While the systems are currently manual, they could be largely automated using technology to get around the bottlenecks. This is what has happened with bank integration: although the European PSD2 (Payment Services Directive) is creating standards and APIs for integrating with bank accounts, this functionality has been provided for many years by companies such as Xero using screen-scraping.
5. Reduced cost.
Once the cost begins to decline — not just the cost of incorporating a company, but making the statutory filings — the potential market will increase substantially. Once this is automated, most likely via SaaS products, and a market is established, usage will increase and costs decrease, as we’ve seen in so many other markets disrupted by code and the internet
6. Artificial Intelligence.
The impact of AI on company formation is potentially vast, probably in ways we can’t imagine. Think about some of the things that AI is already doing — showing us photos of friends based on facial recognition and other signals, making trades based on complex inputs and the underlying lying market environment, playing complex strategic games such as Go. Creating complex networks of companies based on prevailing conditions is surely no more difficult.
7. Techno utopianism.
While the thought of companies created algorithmically may fill some with horror, others will be excited by it, perhaps in the same way people are looking forward to the singularity, when Artificial Intelligence becomes the dominant ‘species’.
The factors outlined above are the core factors that we think will lead to companies being routinely incorporated and dissolved. Note the routinely — it’s certainly possible to do it already in some jurisdictions, and we understand there are already people whose job it is to write computer programmes to create company networks on demand. Just as with the way we stream to music today, or cloud web services, these and other factors will develop and combine, in fits and starts, with new products, services, ideas and even legislation (which will struggle to keep pace with the changes) being created along the way.
Of course this won’t happen in isolation — the impacts on society will be considerable, and societies will react accordingly, at least to the extent that they understand what’s happening, and they have the power and the will to act.
Some will see opportunities, others threats. And very quickly the criminals — increasingly one of the most technologically advanced parts of society, already using software to craft the Crime As A Service industry — will take advantage.
A lesson from history: from paper to algorithmic trading
OpenCorporates has used a Library of Congress photo of the floor of New York Stock Exchange of the 1930s as something of an icon (on business cards, on error pages, and other places) since its launch. It shows a bunch of sharp-looking men in suits in a ‘pit’, buying and selling shares.
Imagine how that felt to be there — the chaos, the noise, the smell.
It’s a far cry from the electronic markets of today, dominated by high-frequency systems and algorithmic trading, black boxes and millisecond advantages, as anyone who’s read Michael Lewis’ Flash Boys will know. Automated trading now accounts for 90% of all stock market trading.
How did we get from one to the other? In the same way we got from analogue telephones and large disconnected computers to the rich web of data-driven apps, websites and services (TV on demand, SaaS, Internet of Things, etc).
Through a series of seemingly independent innovations, failures, pushes from entrepreneurs and their backers, pulls from the market, and changes in both the culture and (less so) the legal landscape.
Today, however, it seems almost inevitable that trading of titles to goods, rights or obligations (which is what the stock markets enable) should be done electronically across networks and via computers using structured data, rather than pieces of paper with handwritten scrawls, filed who knows where, unable to be analysed, and taking an inordinate amount of time to transact and process.
Making the whole system electronic brought a number of benefits:
- Increased liquidity, giving confidence that purchasers will be able to sell their shares, thus giving more confidence to buy
- Reduced costs, both transactional, and in the ‘spread’ between sale and purchase prices
- Introduced more transparency (at least in theory)
- Allowed greater analysis of the market
Once the market becomes electronic, the link with the real world activity is quickly broken. Today almost all activity in the stock market is about the trade, and the expectation of the perceived value of the investment a short distance into the future.
In the derivatives markets this is doubly the case. Originally set up as a way to protect producers (farmers, miners, etc) and consumers (manufacturers) against the problems of price fluctuations, derivatives markets have long outgrown this need, and are now pretty much pure gambling rather than insuring against price fluctuations to give certainty to producers and consumers.
So is this a model for what will happen with company formation activity? Clearly it’s not certain, but we think there are powerful analogues:
- Once upon a time, people bought shares to own a part of a company, and to have a share of the income it would generate. Similarly, not so long ago companies were almost always incorporated to carry on a commercial activity — manufacture a product, deliver a particular service, etc. Now, like derivatives, we’ve moved from the original purpose to a world of special purpose vehicles and flow-through entities, far removed from the underlying activity. Similarly the creation of legal entities to enable money laundering, fraud, or organised crime; to avoid tax, regulation, scrutiny or competition — this is a far cry from allowing entrepreneurs and investors risking capital to fuel innovation and prosperity.Both legal entities and securities (stocks, shares, derivatives etc) are not physical items but legal constructs, allowing them to exist easily in the digital world.
- Both have powerful inbuilt incentives to reducing friction (and hence costs), and increasing opportunities
It’s also worth noting that automated systems don’t mean everything goes smoothly. As anyone with even a passing knowledge of this world knows, these are systems are largely opaque, are not understood, and have serious real-world consequences — flash crashes and systematic instability, short-term thinking and the perverse incentives these produce.
Talking about automated stock market trading, Tom C.W. Lin writes:
This enhanced velocity has shortened the timeline of finance from days to hours, to minutes, to seconds, to nanoseconds. The accelerated velocity means not only faster trade executions but also faster investment turnovers. “At the end of World War II, the average holding period for a stock was four years. By 2000, it was eight months. By 2008, it was two months. And by 2011 it was twenty-two seconds . . . .
But it’s not just the speed of trading, it’s also the decisions taken — about what to trade, when to trade, how to trade, how much to sell or buy for.
It started with the creation of algorithms to split up sales or purchases of securities, to get a better price. The use of algorithms in trading has moved on significantly then — from systems that monitor markets, looking for patterns and opportunities and making trades that take advantages to them, to tools that hack the system by taking advantage of technical quirks to front-run or game it, to the emerging black box strategies of deep-learning driven hedge funds — with algorithms that are not only not public, but unknowable.
Rise of the machines: From manual to automated entities
How far away is this algorithmic company future? There are already many techies looking at this market and, when they see that a company costs less than $20 to incorporate³ asking, not, “Why does it cost so little to register a company?” but, “Why does it cost so much?”
What are the costs apart from the computing time, and the building of the system — which should reduce down to a cents, not dollars, given a reasonable volume, and a lack of human interaction with the process?
And the infrastructure is already being built to enable this.
Want to incorporate in the UK? You can do the whole thing programmatically via the official API. And the latest proposal from the EU is that this facility should be available Europe-wide. Stripe, the payment providers, already has the Atlas programme for making company setup easy, and while there’s no API yet, it certainly seems plausible (disclaimer: Stripe is a client of OpenCorporates).
Further (scary) scenarios
So what does all this mean? How do we make sense of a world where companies — which are, remember, artificial legal constructs created out of thin air to have legal personality — can come into existence for brief periods of time, like fireflies in the night, perform or collaborate on an act, and then disappear? Where there are perhaps not 300 million companies³, but 1 billion, or 10 billion?
When we’re talking about networks of corporations, what if jurisdictions come into existence that are not sovereign states, but platforms, concepts or agreements? When the majority shareholder of, say, a UK company is listed as being XYZ corporation incorporated in ABC jurisdiction. Do we know whether XYZ corporation actually exists in that jurisdiction? Do we actually know that ABC jurisdiction even exists? What if a group of people, or offshore jurisdictions, decide to create or recognise a jurisdiction on a satellite? Could that be stopped, and what are the implications?
Imagine that Amazon decides that — to facilitate trade within a new blockchain platform that it sets up — it will create ‘entities’ in it that can enter into agreements with other entities, all enforced via the underlying rules of the platform. Would that be legal, and how would they, in practical terms, differ from companies today? Arguably they might be more transparent and more trustworthy than IFC companies incorporated in jurisdictions such as the Seychelles or the Cook Islands.
What if the controlling people behind companies are not people at all, but algorithms? Think this is a crazy idea? In fact it’s arguably already possible⁴. What does this mean? Who then is the beneficial owner? Much of the concepts behind company law is predicated on there being humans involved in running and benefiting from companies. But if we have AI or even more limited algorithms making the decisions about what it does, what would these be used for — money laundering, lobbying or funding election candidates, damaging competitors, forming cartels?
(It could be argued that the concentration of power in the big four tech companies — Google, Facebook, Apple, Amazon — is a direct result of the low-friction, data-driven businesses in which they operate, particularly Facebook and Google. Perhaps concentration of power is an inevitable consequence of low-friction, data-driven systems?)
It’s worryingly easy to come up with scary scenarios for how companies will evolve — it’s less easy to know what the implications are, still less to understand how this will impact on society, on the rule of law, on competition, and on the distribution of power in the world.
The next blog in this series, Why the Future of Business Information must be Open, will consider just that – the implications of this explosion of companies, the winners and the losers, and what we can do to tackle it.
Footnotes
- Think about a typical business-to-business transaction, with delivery of goods from one to the other and payment made 30 days after delivery. At the moment, all or some of the following are executed manually:
a. Supplier sends of the invoice to the buyer
b. Buyer enters the invoice in their accounts system
c. Buyer receives the goods, and signs a receipt
d. Buyer checks that the goods are as ordered
e. Some authorised individual at the buyer approves the invoice (having checked that the goods were received and were as ordered)
f. Supplier issues a statement of account
g. Supplier chases up payment of the invoice
h. Buyer pays invoice (eventually, almost certainly not on 30 days)
i. Supplier reconciles the payment received in their bank account with the invoice issued to the buyer
All this could be instead handled by a smart contract, with the condition on which the contract is executed being that the good were received as ordered, in which case a payment is automatically made 30 days after delivery. This massively reduces friction, increases visibility and trust for all parties, and brings other benefits too. For example, not only does failure to pay on 30 days tie up working capital for the supplier, the lack of certainty of when they will actually be paid means they need to have a greater capital buffer to allow for this lack of certainty, meaning the supplier will not use capital as efficiently, and increasing prices for the supplier.
An analogy here might be with Just In Time process in the auto industry, where goods are shipped (and manufactured) with certainty, rather than having large stockpiles to allow for shipping or ordering delays. - It’s just £12 to register a UK company and it can be done entirely programmatically
- Nobody knows how many million companies there are in the world – it depends in part on your definition of a company – but our estimate is somewhere between 250 and 350 million.
- The creation of algorithmic entities is being increasingly discussed in tech and legal circles. It was first proposed by Shawn Bayern in Of Bitcoins, Independently Wealthy Software, And The Zero-member LLC, and critiqued by Matt Scherer in a series of blog posts, who believes that zero-member LLCs are not currently possible, although algorithmically controlled entities using a pair of cross-owned LLCs possibly are. The potential risks of algorithmic entities controlled by AI are discussed by Lynn M. LoPucki. Perhaps, not surprisingly, the discussions are solely around US legal entity forms, although we see no reason why they should appear in the US first, and in fact if the controlling entity for a US company was, say an algorithmic entity incorporated in the Seychelles, for example, no one would even know.
