Multi-tier subcontracting
makes the transition easier.

Japan's multi-tier subcontracting structure is framed as the obstacle to transition. Dissect the structure and the conclusion reverses. Precisely because the structure is multi-tier, the industry shift can proceed without internal lay-offs.

3-06 confirmed that the original driver of IT outsourcing was securing large quantities of coder person-months. This chapter takes the other side: when AI removes coder demand, how the structure built to source those person-months actually moves. The focus is on Japan-specific dynamics, but the conclusion is simple. Multi-tier subcontracting, paradoxically, makes the transition easier.

Multi-tier subcontracting, paradoxically, makes the transition easier

Sketch the typical Japanese SIer structure.

• Prime contractors — large SIers contracting directly with customers, with hundreds to thousands of employees.
• Tier-1 subcontractors — mid-size SIers taking commissioned work from primes.
• Tier-2 and tier-3 subcontractors — smaller vendors and independent contractors below them, supplying head-count by the person-month.
• For some engagements, four to five layers stack.

The reason this hierarchy is disliked is easy to state. Margins pile up at each layer, profit does not reach the working coder, and accountability gets blurred. 3-04 reads it as the structure that inflates price.

But at a transition, the structure's property reverses. Multi-tier subcontracting externalizes coder demand into contracts. Workers are not held inside the prime contractor as employees; they sit in contractual relationships.

What does that mean? When demand disappears, shrinkage happens by not renewing contracts. The prime's own employees do not need to be laid off. Legally, politically, and in public-opinion terms, non-renewal of a contract is far easier than adjusting employment.

Primes can transition without internal lay-offs by not renewing subcontractor contracts

Look at the concrete motion.

Say a prime SIer runs 100,000 person-months a year of engagements. The breakdown looks like this.

As AI-native engagements grow and coder demand falls, what does the prime do?

• Employees stay — dismissal is legally and culturally difficult in Japan. They are reassigned to builder training or to customer-facing judgment work.
• Shrink subcontractor contracts from the bottom up — do not renew tier-3 contracts first, then tighten tier-2 orders.
• New engagements go AI-native — the prime's own employees stand on the judgment side, and AI handles execution.
• Staged shift — keep existing long-term maintenance contracts for their duration, and evaluate AI-native replacement at renewal.

All of this can be executed by central management decision at the prime alone. No internal employment adjustment is required. There are external effects, namely subcontractor downsizing, but that sits inside what the contracts allow.

For that reason, the decision barrier to AI-native transition is low for the prime SIer's management. Conversely, not choosing transition means losing engagements to AI-native competitors. At the point where standing still becomes the higher business risk, the transition accelerates.

Talented subcontractor coders flow to primes, customers, or independence

What happens to the side being shrunk — the subcontractors, especially the coders inside tier-2 and tier-3?

This is hard. When demand disappears, contracts are not renewed. Mid-size subcontractor vendors that stop getting engagements have to scale down or close.

But for talented coders there are several exits.

• Move to a prime — primes need people who can stand on the judgment side, which is to say builder candidates. Coders with judgment ability can be absorbed into the prime's full-time employee slots.
• Move to a customer — hired directly as in-house builders by the customer companies that used to commission SIers (3-06).
• Go independent as an individual contractor or small firm — contracting directly with customers as a builder. This is the management (CIO)-style model from 3-06.
• Move to a different industry — leaving software development is also a path. It is the same kind of redistribution as human computers and typesetters moving to adjacent fields, from 1-03.

Not every mid- or lower-tier coder will move successfully. The layer with judgment ability, or with the willingness to develop it, flows first. This is harsh, but the industry as a whole moves toward talent flowing upward and outward rather than stagnating at the bottom.

The transition rests on labor mobility

Everything above rests on the precondition of labor mobility.

Old-style Japanese employment — lifetime employment, seniority-based advancement, internal reassignment — supported the multi-tier structure. The prime was expected to hold a new graduate as a coder until retirement, and the shortfall was filled with subcontractors. Mobility was low.

That premise has been gradually weakening over a long time. Mid-career hiring has become normal, the recruiting market has matured, and freelance and independent-contractor arrangements have proliferated. The AI-driven industry transition tests the upper bound of that mobility.

If mobility is high, the motion looks like this.

• Mid-career moves between primes increase.
• Prime to customer-company moves grow, which is the builder hiring of 3-06.
• Prime to independence — individual contractor or small firm.
• The subcontractor layers flow into all of the above.

If mobility stays low, it looks like this instead.

• Talent stagnates in shrinking subcontractors.
• New employment forms — treating builders as management, as CIOs — do not develop institutionally.
• The pace of transition is held back by social friction.

Fortunately, mobility is trending upward. The post-COVID spread of remote work, the wider acceptance of side jobs, and the shift toward job-type employment all increase mobility. Society is moving the mobility lever in parallel with the AI shift.

Transitional forms act as shock absorbers

The transition does not happen overnight. It passes through intermediate forms.

• Long-term service contracts — a former coder contracts as an independent with the original employer or a customer company on a long-term basis. Not an employee, but with stable work.
• Secondment and transfer — an SIer employee is seconded to a customer company and becomes a builder there. If results land, permanent transfer becomes an option.
• Internal ventures and spin-offs — an SIer stands up an AI-native business inside, or spins it out as a separate company to avoid conflict with existing engagements.
• Reverse-commission — a customer's in-house builder sells advice to other companies in their domain of expertise. This is the same shape as 1-05's remaining one-tenth of specialists.

These are not permanent structures. They are shock absorbers for the transition. Japanese society has a cultural habit of moving through rapid change by absorbing it into intermediate forms. The multi-tier structure allows rapid shrinkage, and the intermediate forms absorb rapid change. Both effects act in parallel, so the transition is neither abrupt nor stalled.

Physical goods become scarcer than software

The SIer industry's shrinkage is not an isolated labor problem. Over the same few years, several forces are simultaneously pushing total labor demand upward across the economy. They share a single root. The era's scarce resource is flipping from software to physical things.

AI data-center construction is the loudest visible example. The AI boom itself is generating massive demand for physical infrastructure. GPUs, fab equipment, electric power, cooling, buildings, land, networks — every one of them is about physical things, not about code. AI data-center construction is backlogged worldwide, with power supply as the bottleneck. The cheaper AI becomes, the scarcer the physical things that run AI. This is the most visible sign of the era in which physical goods get scarcer than software.

Reshoring of manufacturing pushes the same way. Middle East instability, geopolitical energy-price rises, and global logistics-cost increases are eroding the economics of offshore production. Domestic manufacturing in Japan — especially high-value, low-volume, fast-response work — gains relative competitiveness. As reshoring proceeds, demand for shop-floor labor, production design, and manufacturing engineering rises.

The shift to natural farming stacks on top of that. The supply of the major chemical-fertilizer inputs — ammonia synthesized from natural gas, potash from Russia and Belarus, imported phosphate rock — is destabilizing, and prices are rising. Once the input cost of chemical agriculture crosses the breakeven line, natural farming is no longer a choice but a necessity. Natural farming requires more labor than chemical agriculture across soil preparation, weeding, and harvest, so agricultural labor demand also moves upward (the structural details are covered in the separate aiseed.dev series Phosphorus Depletion and Natural Farming).

The three forces — physical-infrastructure demand from AI itself, reshoring of manufacturing, and the shift to natural farming — run on the same time scale as the SIer industry's shrinkage. As a result, the options open to displaced coders broaden significantly. Alongside intra-industry flow to primes, customers and independence, out-of-industry physical labor demand becomes a major channel.

The historical parallel from 1-03 — human computers and typesetters moving to adjacent fields — was viable only because labor demand happened to exist where they landed. The same applies here. The side where labor demand disappears, code production, and the sides where it grows, physical production, agriculture and AI physical infrastructure, are moving in parallel inside the same society.

Mobility rises over time

Confirm the direction of the transition period.

Labor mobility, job-type employment, treating builders as management in the CIO seat, and social acceptance of business commissions have all measurably risen over recent years. Multiple forces will keep them rising.

• Demographics — a shrinking working-age population pushes mobility up on pure economic grounds.
• International comparison pressure — comparison with overseas executive and management markets, especially the US, pushes Japanese companies to revise compensation.
• AI itself as pressure — talent that does not fit old-style employment models accumulates: builders who should be brought in as management and CIOs, and the AI that handles the professional execution. That accumulation creates pressure for institutional reform.
• Policy direction — government policy on job-type employment, legalized side jobs, and advanced-professional regimes all point the same way.

Mobility will keep rising. As time passes, the friction of the move into an AI-native industry structure shrinks. The friction maximum is in the first few years, and after that the transition accelerates.

The education and hiring axes move at the same time

Alongside labor mobility, another axis has to move: the foundational discipline of the technical profession. Japan's science-and-engineering education has long centered on programming languages, frameworks, and design patterns — the core of software engineering. Once AI has taken that core, the human side has to shift its weight onto the liberal arts (1-04): logic, verbalization, ethics, systems thinking, history.

The shift runs through everything from university curricula to corporate hiring criteria. From "can you write code?" to "can you judge?" The question changes on the education side and the employment side at once.

Summary

Japan's multi-tier subcontracting structure, paradoxically, makes the transition easier.

• Structure — multi-tier subcontracting externalizes coder demand into contracts. When demand disappears, shrinkage happens by not renewing contracts
• Primes — can transition on a central management decision alone, with no internal employment adjustment
• Subcontractors — this is hard, but the layer with judgment ability flows to primes, to customer companies, and to independence
• Precondition — the transition rests on labor mobility, and mobility is trending upward
• Transitional forms — long-term service contracts, secondment, internal ventures, and reverse-commission act as shock absorbers
• Outside the industry — AI physical infrastructure, manufacturing reshoring, and the shift to natural farming open labor demand on the same time scale
• Education and hiring — the foundational discipline moves from software engineering to the liberal arts

The structure of the industry and the way people move through it are now in view. What remains is where the transition actually starts. Not from a prime contractor's management decision, and not from government policy.

The next chapter takes up the case that the AI revolution starts from below: that it begins wherever no approval process is needed, that top-down adoption leaves the company with nothing, and what stays in the company when it starts from below. How long the transition takes to complete is treated in the final chapter, 3-09.

Related articles

• 1-03: AI Now Does the Software Engineer's Work
• 3-04: The Structural Uneconomy of the SIer Model
• 3-06: Companies Hire Builders
• 3-08: The AI Revolution Starts From Below
• Phosphorus Depletion and Natural Farming
• Structural analysis 08: Subtracting the enterprise-IT tax
• Structural analysis 12: AI and the sole proprietor