The modern company wasn’t designed. It was improvised — a series of workarounds for three problems that dominated business in the late 19th century: slow information, limited human cognitive bandwidth, and expensive coordination.
Agentic AI eliminates all three. Which means the org chart that Taiwan’s bicycle industry has run on for the past 40 years isn’t just inefficient anymore.
It’s a solution to problems that no longer exist.
Contents
The Problem
The modern company structure wasn’t designed. It was improvised to workaround three specific problems that dominated business in the late 19th century.
Problem one
In 1880, information moved slowly.
Coordinating people across a complex task required supervisors physically present to observe, assess, and redirect.
Every management layer was a signal relay station involving the gathering of information from below, filtering it, then passing decisions downward.
The hierarchy existed because information couldn’t move faster than a human carrying it.
Problem Two
Human cognitive bandwidth is limited. No single person can actively coordinate 200 workers.
So companies broke coordination into chunks that fit inside individual human minds: departments, divisions, and reporting lines.
In other words, the entire org chart is a structural response to a biological ceiling.
Problem Three
Coordination through markets was expensive. This is Ronald Coase’s Nobel Prize-winning insight from 1937.
Firms exist because finding people, negotiating terms, and managing quality in open markets carries real costs.
When those transaction costs are high, it’s cheaper to bring everything inside a firm and manage it through hierarchy.
The company, in Coase’s framing, is a transaction cost reduction machine.
Three constraints; three workarounds; one org chart that has persisted largely unchanged for 150 years, not because it’s optimal, but because nothing eliminated the underlying constraints.
Until now—and the rise of the AI Agentic economy
Agents
Agentic AI attacks all three simultaneously. Information transmission is instantaneous and effectively free.
Agent systems don’t have cognitive bandwidth ceilings. One person overseeing a well-configured agent stack can manage operational complexity that previously required a mid-sized management layer.
And the transaction costs Coase identified as the reason firms exist in the first place are now a fraction of what they were, which changes the calculus on what needs to be inside a firm at all.
The traditional firm then then becomes a unit comprised of a small human centre, holding judgment, relationships, and accountability, and surrounded by configurable agent capacity assembled around specific goals rather than maintained as permanent headcount.
The org chart was never a reflection of how business works best. It was an engineering solution to problems that no longer exist in the same form.
The companies that see this clearly, and rebuild around what’s actually possible now, won’t just outperform competitors still carrying legacy overhead.
They’ll operate in a different economic reality entirely.
The world of slow information, human bandwidth ceilings, and expensive coordination has gone.
What remains is the organizational response to it, still running on inertia.
The consequences haven’t fully arrived yet.
But that’s a description of lag, not of the underlying reality.
Why The Bike Industry? Why Now?
Taiwan’s bicycle sector is one of the most instructive places to examine what the agentic AI transition actually means for real businesses because it’s a well-defined case study.
Clear structure. Recognizable company types. Publicly traded anchor companies. A shared organizational logic that has changed very little since the sector matured in the 1980s and 1990s.
It also has a specific problem, the one all of us in the industry recognize oh so well.
After years of post-pandemic boom followed by an equally steep correction, most Taiwan bicycle companies are carrying legacy overhead into a margin-compressed, demand-uncertain environment.
Structural pressure from outside. Organizational rigidity from inside. And that outside pressure just got measurably worse.
China’s 15th Five-Year Plan targets 90% AI integration across its economy by 2030, not as an aspiration but as a state infrastructure project, backed by the same subsidy architecture that drove Chinese solar and EV manufacturing to global dominance.
In 2024, China installed 295,000 industrial robots, 54% of the global total, at a density that now surpasses Germany and Japan.
Meanwhile, a March 2026 survey by CPA Australia found that 33% of Taiwan SMEs identified AI as their top technology investment priority. Intent, not capability. A 33% interest rate against a 90% state mandate.
Taiwan’s Silicon Shield protects borders. It does not protect balance sheets. The 80% of Taiwan’s workforce employed in SMEs are not in the semiconductor sector that generates the headlines. They are in exactly the kind of mid-market manufacturing that China’s AI industrialization is systematically targeting.
The threat to bicycle component manufacturers isn’t immediate in the sense of next quarter’s price list. It’s structural, compounding, and moving in one direction.
And transitioning an SME from its current operating model to one built around agentic capability takes 24 to 36 months at minimum.
Companies that wait for the threat to become visible on a competitor’s quote sheet are already making a decision — they just haven’t acknowledged it yet.
That combination — legacy overhead, margin compression, and an accelerating external cost-floor reset — is exactly where the agentic transition becomes impossible to ignore.
The argument is not that Taiwan’s bicycle companies are doing anything wrong. It’s that they are organized correctly for a world that no longer exists.
The Structure of the Industry
Taiwan’s bicycle ecosystem operates across four tiers, each with its own organizational logic and its own exposure to the coming transition.
Tier One: The Behemoths
Giant Manufacturing and Merida Industry define the industry’s upper boundary.
Both are vertically integrated, publicly traded, and run organizational structures that would be immediately recognizable to any corporate manager anywhere in the world: functional departments, regional subsidiaries, dedicated marketing and export teams, multi-layer management hierarchies.
Giant operates a bunch of manufacturing factories across Taiwan, China, the Netherlands, and Hungary, managing several consumer brands (Giant, Liv, Momentum, CADEX).
In 2024, women represented some 30% of management roles and the company’s internal job rotation rate reached 63%. Those are metrics that only appear in companies with substantial HR infrastructure — which is itself a significant coordination overhead.
Merida manufactures over two million bicycles a year and markets in over 77 countries, with R&D headquarters in Germany. The organizational cost of operating at that scale is substantial.
Tier Two: The Specialist Manufacturers
Companies like Pacific Cycles (Taoyuan, founded 1980) and KMC Chain (Tainan, founded 1977) are globally significant in their niches, but operating at a scale where organizational overhead is felt directly in margins.
Pacific provides ODM services for over 40 bicycle brands worldwide from its Section Zero R&D facility, with more than 20 engineers dedicated to product development.
KMC operates 11 production plants and 16 distribution entities across Taiwan, China, the Netherlands, the US, Indonesia, and Vietnam, serving over 150 countries.
Sixteen distribution entities means sixteen sets of management infrastructure, each one built because there was no other way to do it.
Tier Three: The Component Specialists
Tier three consists of hundreds of mid-sized Taichung and Changhua County manufacturers — bottom bracket makers, hub manufacturers, crankset producers, saddle companies — operating predominantly in the 50 to 500 employee range. FIRST sits squarely in this group.
These are the backbone of the Taiwan supply chain. Most have minimal English-language marketing capability, limited digital presence, and organizational structures built around a founder or family ownership model with functional departments added as the company grew.
Tier Four: The Micro-Specialists
Small component and accessory manufacturers, often family-owned, 10 to 50 employees, serving niche applications or specific OEM relationships characterize tier four.
They have the thinnest overhead, the most flexibility yet the fewest resources to navigate structural change. But as we’ll see, some of these are paradoxically the best positioned of all.
Where the Org Chart Overhead Actually Lives
The 150-year-old bug manifests differently across these tiers, but it’s present in all of them.
At Giant and Merida, the overhead is visible and documented.
The HR infrastructure, regional management layers, export coordination teams, supplier relationship managers — each exists because information movement and execution coordination still run through human intermediaries at scale.
Each is a candidate for agent substitution.
At Pacific Cycles, the picture is more nuanced. The company is currently building its third factory in Taoyuan, investing in manufacturing capacity at precisely the moment the organizational question is becoming urgent.
Pacific has always been design-led and founder-driven. George Lin’s philosophy of innovation over volume is embedded in the company’s DNA.
That design intelligence is exactly what agents cannot replicate.
But the administrative and coordination overhead surrounding that core intelligence—export documentation, distributor management, product catalogue maintenance, customer inquiry handling etc.—is a different matter entirely.
The Functions Most Immediately Exposed
Across all four tiers, the same functional categories emerge as most vulnerable. None of this requires speculative future technology. The capability exists today.
Export marketing and content production.
Most Taiwan component manufacturers employ one to three people managing websites, catalogues, trade show materials, and distributor communications.
Functions such content creation, translation, catalogue management, specification documentation and so on, is almost entirely executable by current-generation AI agents with a single human editor in oversight.
The companies spending on small marketing teams for this purpose are paying for a function that costs a fraction as much to perform at comparable quality.
Customer inquiry and pre-sales qualification
A significant proportion of communication between Taiwan manufacturers and international buyers consists of repeated, formulaic exchanges: MOQ questions, specification clarifications, lead time inquiries, sample requests, price negotiation within known parameters.
These conversations happen over email, WeChat, and LINE. The human sales staff performing this function are mostly acting as information relays. Not relationship builders. Not strategic decision-makers. Information relays.
Distributor and channel management
Coordinating orders, managing complaints, sending promotional materials, tracking inventory levels at distributors are largely schedulable, rule-based, and information-relay in nature.
In Coasian terms, it’s a transaction cost reduction mechanism. Agents will replicate it at a fraction of the cost.
Market research and competitive intelligence
Taiwan’s export-oriented manufacturers operate with limited real-time intelligence about their target markets.
The information exists in trade publications, distributor sites, show coverage, and social channels.
Synthesizing it currently requires human researchers, or simply doesn’t happen systematically. Agent systems can do this continuously at negligible cost.
Huge things happening in relation to data—it’s key. Check out Bicycle Cluster’s Data Governance project.
The Competitive Exposure by Tier
The risk isn’t uniform — and that matters for how this argument lands with different audiences.
Giant and Merida are not at existential risk. They have the capital, the management capacity, and the global relationships to navigate the transition. Their risk is different: they move slowly by necessity, and the disruption will come from below.
A new mid-range brand entering their addressable market, built on an agent-native operational model with a fraction of the overhead, can undercut their price points while matching their marketing quality. The giants will adapt, but later and more painfully than they need to.
The specialist manufacturers—KMC, Pacific Cycles, and their equivalents—are in the most interesting position. They have genuine competitive moats in product and design capability that agents cannot replicate.
KMC’s chain engineering knowledge, Pacific’s folding bike design intelligence, are irreplaceable. But both carry organizational infrastructure built for a world where coordinating global operations required human intermediaries at every node.
That infrastructure will become progressively more expensive relative to agent-native competitors who can match their sales and marketing output with a smaller team.
Looking at the rate of change just since February (OpenClaw and the very recent Nvidia answer to the threat from the edge, NeMo Claw), what is the window to restructure proactively?
Three to five years? A month ago I would thought so.
Now? Today…? I feel the window continuing to shorten.
The component specialists — the mid-sized Taichung and Changhua manufacturers — arguably face the most acute near-term pressure and have the least capacity to respond.
A founder with fifty employees, two of whom handle export marketing and three of whom manage customer inquiries and order coordination, is paying for five human signal-relay functions that could be handled by a well-configured agent stack supervised by one person.
That is not a marginal efficiency gain. For a company operating on single-digit margins in a down market, it is a survival variable.
The micro-specialists are the wild card.
Some will be absorbed into larger structures. Some will simply be outcompeted.
A few, paradoxically, are best positioned of all: founders who are the entire company, with deep craft expertise, minimal overhead, and enough agility to build an agent-assisted operation around their core skill.
The artisan manufacturer with one human and a well-configured agent stack is not a fantasy.
It’s already happening in other verticals at a pace that accelerates daily.
The Merida Paradox
Merida’s stated management philosophy is built around empowerment and flat structure.
Employees have space to implement their own ideas without strict one-dimensional performance metrics.
Here is the paradox.
The value of that philosophy is that it gives employees space to exercise judgment, creativity, and initiative — precisely the functions that agents cannot perform.
The flat structure argument is, at its core, an argument for concentrating human energy on high-judgment work and reducing bureaucratic coordination overhead.
That is exactly what the agentic transition argues for.
Merida’s own management philosophy points toward the architecture that agentic AI makes possible.
The question is whether the company arrives at that architecture through deliberate choice or gets pushed toward it by competitive pressure.
Those are very different journeys.
What Comes Next?
The transition will not arrive as a single disruptive moment.
It will arrive as a series of competitive disadvantages that accumulate into a structural crisis for companies that don’t respond.
The first signal will be a new market entrant, unlikely from Taiwan, possibly from a European or North American brand with a lean team and an agent-native operation.
They will achieve marketing quality, customer responsiveness, and product development speed at a cost structure that established Taiwan companies cannot match while maintaining legacy overhead.
The second signal will be a Taiwan company that restructures proactively and demonstrates the cost and output profile that becomes the new competitive benchmark.
The third signal will be the trade show floor. Eurobike is already contracting. Taipei Cycle is holding, but the pattern of attrition is visible.
How many of those missing companies are quietly replacing show presence with digital-first, agent-assisted outreach that costs a fraction as much and reaches buyers more directly?
Hardly any yet.
But the ones that have/are in the process of implementing will not return to the old model.
This is not an argument for Taiwan’s bicycle companies to become technology companies.
The expertise, the relationships, the manufacturing capability, the product intelligence — those are the business.
The org chart is a 150-year-old coordination mechanism, built for constraints that no longer exist.
It was never the business. It was the scaffolding around it.
And the scaffolding is due for removal. The agents are coming to transform how those relationships are expressed in the rapidly approaching future.
The companies that recognize the distinction — and act on it before the competitive signals become impossible to ignore — have a genuine window.
Do I have to point out that It won’t stay open indefinitely? 😄
The transition from structural threat to operational reality doesn’t announce itself.
It just arrives, and the gap between where you are and where you need to be comes an an unbridgeable chasm.
🫵 “You’re on the menu!”
A 50+ year veteran of the bicycle industry said it after the pandemic and again, most recently at Taipei Cycle’s Industry Forum.
He wasn’t talking about the rise of the Agents or anything AI.
Let’s steal the phrase and venture that yes, indeed—anyone not going full send into learning how the new world works is absolutely on the menu.
EN 
DE