Who owned the horses?

A report by Simon Brender, 2026.

Prologue

We treat the modern company as the strongest thing in its market. A big firm run on industrial-era logic: own the assets, hire the people, scale the output. The evidence was everywhere, in firms with payrolls the size of small countries.

But this arrangement is less permanent than we assume.

The reasons that once justified the full-time employee are fading. Coordination was expensive. Information was expensive to move. Physical assets could not be moved at all. When those reasons go, the structures built on top of them go too.

To see where this is heading, start with the horses.

1. Why organisations hired people in the first place

Economics gave us the firm.

Ronald Coase (1937) argued that companies exist to cut transaction costs: the friction of constantly contracting for skills and services on the open market.

Oliver Williamson took this further. Firms bring an activity in-house when the outside market is too slow, too risky or too uncertain to rely on.

So organisations added headcount for four reasons. Steady access to skills was cheaper than negotiating each time. Coordination needed people in the same place. Knowledge was hard to move. And capital and production were tied together.

Firms became dense networks. They pulled the nodes, the workers, machines and intellectual property, inside the boundary to cut friction at the edges.

The industrial revolution poured fuel on that.

2. Horses, machines and the first workforce automation cycle

Before mechanisation, horses were a contractible resource. They were often owned independently and hired into farms, mills and transport operations.

As machinery spread, organisations brought horsepower in-house. Owned power was more predictable, more scalable and easier to control.

There was a network effect. Work centred on factories because machinery made for heavy, fixed nodes. Labour moved to where the capital was, not the other way round.

Adoption was not smooth. Early adopters gained productivity and pulled talent and capital inward. Late adopters lost pricing power and disappeared. Some sectors kept horses far longer because the infrastructure and the trust networks lagged behind.

This was a rewiring of the network, from scattered agrarian clusters to hub-and-spoke capitalism.

3. Minds over muscle: the knowledge economy's second rewire

Move on a century, and the factories ran on a new input: human thinking.

After the Second World War, as information spread, first Xerox copiers, then ARPANET, the frictions shifted. Knowledge that had been tacit and tied to a location became easier to write down. But organisations kept adding headcount, not because machines had failed, but because minds were now the scarce resource.

Economics adapted with them.

Gary Becker, in his 1960s human capital theory, argued that workers were not interchangeable parts. They were investments, trained, specialised and loyal. Firms brought talent in-house to capture the return on that capital.

Michael Jensen and William Meckling, in their agency theory of 1976, showed how misaligned incentives in outside markets, for instance freelancers chasing short-term work, eroded trust. The answer was to bring the work in-house for alignment.

So the white-collar boom followed. Specialised skills scaled through hierarchies: R&D labs, consultancies and tech campuses as idea foundries. Information asymmetry flipped, with firms hoarding data, customer insight and proprietary models, to stay ahead of spot markets. And network effects amplified it, as talent clusters like Silicon Valley pulled whole ecosystems inward and cut search costs.

But this was not permanent. Gig platforms such as Upwork and Fiverr cut negotiation frictions by 80 per cent in creative sectors (McKinsey, 2023). Remote tools like Slack and Notion separated proximity from coordination. And AI began to suggest that minds, too, could be orchestrated rather than owned outright.

The rewire ran from rigid org charts to fluid talent pools. Early adopters gained agility: Netflix's no-VP keeper test released iterations 40 per cent faster. Laggards carried bloated payrolls into rising labour costs, up 5.2 per cent year on year globally (ILO, 2024). Some sectors held on to full-time longer, in particular infrastructure-heavy fields like energy, where trust lagged the digital twins.

The network shape changed too. Hub-and-spoke became a mesh, with the fractional experts linking through protocols rather than bosses.

4. The fragility of giants

Few companies survive for centuries. Here is why.

Once a firm becomes a central node, one the network cannot route around, it gets three advantages: better information, first pick of capital, and a pull on talent.

That same centrality makes it rigid. Coordination gets harder faster than the firm grows. Innovation slows. The nodes at the edge move quicker. And central nodes tend to be disrupted from those edges.

Kodak was not stupid. It was highly central in a network that changed shape under its feet.

The lifespan of an S&P 500 company has fallen from about 75 years in the 1950s to under 15 years today (McKinsey).

The cause is easy to miss: the network rewires faster than the organisation can.

5. Synthetic cognition: the third cycle

Now AI arrives, not as a tool, but as the thing that removes the friction.

The first cycle was about owning muscle, from horses to steam. The second was about employing minds. This one is about orchestrating synthetic cognition: intelligence supplied on demand, at low cost, without a fixed limit.

Look at the forces that are falling away.

Transaction costs fall to near zero. Large language models such as Grok-4 negotiate skills in milliseconds, with no contracts, just APIs.

Coordination becomes near-instant. Agentic systems, for example multi-agent RLHF stacks, align themselves without human go-betweens.

Knowledge transfer becomes near-perfect. Digital twins mirror expertise and scale it without diluting it.

Capital and output come apart. Cloud infrastructure and open models let almost anyone orchestrate at marginal cost.

Firms will not vanish. They will spread. But the reason they exist is the casualty. Coase's bargain breaks: why bring work in-house when external synthetic cognition is faster, cheaper and more certain? Williamson's risks get hedged by probabilistic forecasting, with Bayesian agents outperforming human teams by 25 per cent in simulations (DeepMind, 2024).

This shows up in practice. Celerio's model of fractional orchestration cuts coordination overhead by 40 per cent by layering AI twins over expert nodes. The result is market traction without the weight: scale through intent, not headcount.

Adoption comes in waves. Pioneers orchestrate cognition directly. Mid-tier players like consultancies hybridise, and risk inertia. Laggards in regulated sectors such as finance and pharma delay, but trust networks, for instance blockchain provenance, will force the switch.

This is reinvention, not replacement. The firm directs external AI and experts towards an outcome instead of housing them. The early pilots show 3x return on the effort put in, with B2B startups reaching product-market fit six months faster through orchestrated go-to-markets.

Is it smooth? No. Ethical guardrails lag, whether that is bias in synthetic decisions or cascades of hallucination. We will iterate, as always, treating the constraints as things to optimise rather than dodge.

6. Three cohorts under pressure

A) Startups: innovation stagflation

SVB's collapse showed how concentrated startup liquidity had become. At the same time, AI coding agents produce fast gains at the level of the individual company, but they degrade the signal investors rely on: headcount no longer stands in for value.

Capital shifts to fewer, better-connected winners. Innovation narrows. The long tail stagnates.

B) Top-tier consultants: the high-trust nodes

Research and synthesis? AI does that now, and fast. But judgement, alignment, managing the politics and accountability stay largely human.

So the Tier 1 consultancies shift. They become orchestrators. They own platforms, meaning the intellectual property, the tools and the fabric. They govern networks rather than only advising. Their moat moves from owning knowledge to sitting at the centre of relationships.

C) Financial services: automation with systemic risk

Banks automated long ago. AI now moves into risk models, AML and KYC, surveillance, personalised pricing, and compliance and reporting.

The result is lower friction, higher concentration and faster contagion. Regulators worry about crises that hit at machine speed. Finance stays central, but it is brittle.

7. Where this is heading

We are watching a structural inversion. The line marking the edge of the firm, once solid, turns into a dotted line.

Organisations will look less like pyramids. Expect a small core, a wide distributed edge, and AI linking the two. The ecosystem matters more than the enterprise.

8. What survives

What stays scarce becomes valuable. Four things will.

First, being modular, with replaceable parts and a persistent identity. Second, owning mission-critical trust: governance, safety and relationships. Third, making markets for capability rather than hoarding labour. Fourth, adapting faster than the network around you rewires.

The next century of giants may not be the firms with the most employees. It may be the ones with the most influence at the edges.

What leaders can do now

For leaders thinking ahead, four steps.

Map your coordination costs. Work out where coordination drags, and which of those costs you could meet from outside the firm rather than inside it.

Run a small pilot. Start with fractional AI layers on core teams, and measure the traction you gain per unit of effort you put in.

Keep an audit trail. Invest in provenance, such as auditable agent logs, to close the trust gap. That is your edge in B2B Asia, where Singapore's mix of cultures speeds adoption.

Expect to get it wrong first, and expect the payoff to still be large. Be honest about it, and iterate faster.

The future is not about hiring more. It is about activating capability when you need it, through fractional talent, digital twins and intelligent systems that work continuously.

Labour costs rise faster than productivity. Reliance on consultants grows. AI removes friction. Three pressures, one opportunity.

Final thought

So who owns the horses now? No one. That is the point.

The last automation cycle retired the working horse. This one may do the same to the full-time employee. AI does not replace the organisation. It replaces the reason the organisation exists.

This is not the end of work. It is the end of owning the workers.

We help forward-thinking organisations work through this shift. Explore our approach. Read our latest insights.