AI-driven development is beginning to unlock margin leverage for Japanese IT services companies

Overview

This report focuses on Japan's IT and software industry, especially how large system integrators and IT services companies can improve efficiency across the full software development lifecycle through AI-driven development, AI-native development, and multi-AI-agent platforms. The report argues that the industry has long relied on multilayer outsourcing structures, with outsourcing costs becoming the largest cost item; if leading contractors can bring outsourced development, testing, operations, and maintenance back in-house and shift their business model from "person-month × labor hours" to value-based pricing, margins could improve structurally.

Core views

The core views are: first, AI is no longer just an assistive tool, but is entering the full development lifecycle including requirements definition, design, coding, testing, and operations/maintenance; second, in Japan's information services industry, outsourcing costs as a share of revenue rose from about 20% in FY1991 to a weighted average of 33.5% in FY2024, above the 27.3% labor-cost ratio in the same year, so lower outsourcing costs offer significant margin leverage; third, the cases of Fujitsu, NEC, NRI, and NTT DATA show that productivity improvement has moved from proof of concept to business deployment; fourth, industry differentiation will depend on whether companies can overcome bottlenecks such as data foundations, process redesign, AI talent, governance reliability, and organizational inertia.

Analysis framework

The report combines industry cost-structure analysis, corporate case studies, and an AI deployment maturity framework. On the cost side, it focuses on outsourcing costs as a share of revenue, personnel costs as a share of revenue, and multilayer subcontracting structures; on the corporate side, it compares the AI platforms, productivity targets, deployment pace, and business-model adjustments of Fujitsu, NEC, NRI, and NTT DATA; on the demand side, it cites survey findings from McKinsey, Deloitte, Gartner, BCG, PwC, and others on enterprise AI adoption, ROI, data foundations, and organizational transformation.

Methodology notes

• industry cost structureoutsourcing-cost margin leverage

  the transmission of lower outsourcing costs to operating margins

  The report views outsourcing costs as the largest optimizable cost pool for IT services companies, arguing that if AI-driven development enables prime contractors to reduce dependence on multilayer subcontracting, operating margins could rise by about 5-10 percentage points from the current 10%-20% range.

• business model transformationfrom person-month pricing to value-based pricing

  the pricing model determines retention of efficiency gains

  If efficiency gains are fully passed through to customers via price cuts, margin improvement will be limited; only by shifting from hourly billing to charging based on delivery value, service value, or usage can companies more likely retain the productivity gains brought by AI.

• organizational transformationfrom pilots to scaled deployment

  bottlenecks in AI implementation

  The report emphasizes that many enterprises use AI frequently but still remain at the pilot stage; insufficient data foundations, unrestructured business processes, shortages of AI talent, governance reliability, and Japan-specific legacy system issues are the main barriers to scaled deployment.

Asset mapping & comparison

Structured mapping from thesis to named assets (strengths, weaknesses, peers, risks).

• Fujitsu (6702.T)

  core case company

  Strengths

  It has the AI-Driven Software Development Platform and the proprietary LLM Takane, demonstrated a 100x productivity improvement case, and has clearly planned a transition toward service-value pricing.

  Weaknesses

  It needs to extend proof-of-concept results to more industries and customer scenarios, while bearing the costs of platform rollout and organizational transformation.

  Comparison

  Among the SI companies cited in the report, Fujitsu's case places greater emphasis on multi-AI-agent platforms, quality-control mechanisms, and pricing-model change.

  Risks

  Customers may demand a share of cost savings, platform deployment may fall short of expectations, and upfront AI investment may pressure short-term profits.

• NEC (6701.T)

  core case company

  Strengths

  It proposes becoming an AI Native Company, applying AX to about 30 BluStellar scenarios, and supports enterprise AI deployment through cotomi and AI Platform Service.

  Weaknesses

  It needs to prove that revenue growth, incremental profits, and SG&A improvement from BluStellar scenarios can be sustained.

  Comparison

  NEC places more emphasis on margin improvement driven by consulting-led engagement, scenario-based services, and group/partner collaboration.

  Risks

  Scenario commercialization may be slower than expected, AI platform competition may intensify, and higher customer requirements for governance and private deployment may raise delivery complexity.

• Nomura Research Institute (4307.T)

  core case company

  Strengths

  It uses the NRI Enterprise AI-Driven Development Model as the core of production innovation, plans to achieve a 100% adoption rate for AI-driven development by FY2028, and is advancing a financial-services-specific LLM.

  Weaknesses

  The transformation needs to expand from financial services experience to broader system integration businesses, creating high execution complexity.

  Comparison

  NRI places greater emphasis on a phased roadmap, moving from downstream development efficiency gains in 2025-2026 toward AI-first development model restructuring in 2026-2028.

  Risks

  The pace of industry expansion, customers' willingness to redesign processes, and financial-grade reliability requirements may affect the rollout pace.

• NTT DATA

  core case company

  Strengths

  It has already begun practical deployment in 2025, achieving a 70% reduction in labor hours and about 3x productivity improvement in web application development, and has launched LITRON Builder.

  Weaknesses

  It needs to replicate these results across a large number of projects while also addressing differences in governance, security, and usage environments among large enterprises.

  Comparison

  NTT DATA's path places more emphasis on phased maturity expansion and the long-term goal of a 70% overall process reduction by 2030.

  Risks

  The pace of automation expansion, customer security and governance constraints, and the risk of missing sales targets for generative AI-related business.

• Japan IT services and system integration industry

  theme beneficiary sector

  Strengths

  It has a large outsourcing cost pool, a long development lifecycle, and strong demand for legacy-system transformation, giving AI-driven efficiency gains meaningful operating leverage.

  Weaknesses

  Multilayer subcontracting structures, low tolerance for failure, weak data foundations, and organizational inertia make transformation difficult.

  Comparison

  Compared with general software tools companies, Japanese SI companies have more opportunity from cost-structure optimization and delivery-model restructuring.

  Risks

  Efficiency gains may be absorbed by customer price pressure, payback periods on AI investment may lengthen, and insufficient governance reliability may hinder scaled deployment.

Key data

• outsourcing costs as a share of sales31%-40%The report states that outsourcing costs for prime contractors amount to 31%-40% of sales, making them an important source of margin improvement.
• potential operating margin improvementabout 5-10 percentage pointsEstimated upside from structurally lower outsourcing costs and higher development efficiency.
• Japan information services industry FY2024 outsourcing cost/revenue ratio33.5%Weighted-average basis, higher than the personnel cost/revenue ratio in the same year.
• Japan information services industry FY2024 personnel cost/revenue ratio27.3%More than 6 percentage points lower than the outsourcing cost ratio.
• Fujitsu productivity casesystem modification completed in 4 hours versus about 480 hours previously requiring 3 person-monthsA proof of concept related to medical fee revisions, showing about 100x productivity improvement.
• NRI productivity case10-30xAchieved in some financial ASP cases at the detailed design, development, and unit testing stages.
• NTT DATA medium- to long-term target70% overall reduction in development processes by 2030The company plans to use generative AI in 500 projects in FY2025 and achieve 20% productivity improvement, then raise the share of AI projects to 50% in FY2027 and achieve 40% productivity improvement.
• enterprise AI maturityonly 1% have reached the mature stage, and 19% have proven ROIThe report cites surveys including Deloitte showing a significant gap from AI awareness to scaled deployment.
• Japan AI deployment gap8.9% enterprise deployment rateThe report cites PwC saying Japan ranks last in a five-country comparison; only 13% of Japanese companies said AI results exceeded expectations, versus 51% in the U.S.

Impact & implications

The investment implication is that large Japanese SI and IT services companies may see an AI-driven margin re-rating opportunity. Companies with proprietary or controllable AI platforms, strong industry knowledge, customer process redesign capabilities, and value-pricing capabilities are more likely to convert efficiency gains into profits; by contrast, if companies remain constrained by person-month billing, customer price pressure, subcontracting inertia, and legacy systems, productivity gains may be difficult to fully reflect in earnings.

Risks

• Customers may require IT service providers to pass through AI-driven cost savings via price cuts, weakening margin improvement.
• AI platforms, data foundations, talent, and process redesign require upfront investment, which may pressure profits in the short term.
• Organizational inertia and Japan's multilayer subcontracting structure may slow insourcing and the transformation of development models.
• Legacy systems, reliability, and governance requirements may limit the scaled application of AI agents in critical system development.
• Many enterprises are still at the AI pilot stage, with insufficient proof of ROI, so there remains uncertainty over whether industry demand will shift from pilots to scaled deployment.
• Morgan Stanley discloses investment banking, service, or potential conflict-of-interest relationships with multiple covered companies, and investors should consider research independence risk.

What to watch

• Whether Fujitsu can expand its multi-AI-agent platform from healthcare and public-sector contracts to industries such as finance, manufacturing, and distribution, and implement value-based pricing.
• Whether NEC BluStellar scenarios deliver verifiable revenue growth, incremental profits, and SG&A efficiency improvement.
• Whether NRI achieves a 100% adoption rate for AI-driven development by FY2028 according to its roadmap, and deploys fully autonomous AI agents by FY2030.
• Whether NTT DATA completes its 500 generative AI projects in FY2025, reaches a 50% AI project share in FY2027, and achieves the 70% development process reduction target by FY2030.
• Whether customer contracts shift from person-month billing to value-, outcome-, or usage-based pricing.
• Whether the outsourcing cost/revenue ratio and operating margins show sustainable improvement in financial statements.