Contact

Contact HaxiTAG for enterprise services, consulting, and product trials.

Showing posts with label Enterprise Al solutions. Show all posts

Wednesday, May 6, 2026

CyberAgent's Enterprise-Level AI Agent Deployment: Unpacking the 93% Active User Rate Through Voluntary Adoption Strategy

May 06, 2026

Case Overview and Core Themes

Company Background and AI Strategic Framework

CyberAgent, a leading Japanese internet company with diversified business operations spanning advertising, media and IP, as well as gaming sectors, stands as a representative enterprise in the Asia-Pacific technology, media, and entertainment industries. The company's journey into artificial intelligence began as early as 2016, when it established a dedicated AI laboratory (AI Lab) focused on AI research and development related to digital marketing. This early strategic investment laid a solid technical foundation and cultivated an organizational culture that would later prove instrumental in the successful deployment of enterprise-level AI agents.

In 2020, CyberAgent launched the "Kiwami Prediction AI" system, specifically designed for intelligent optimization of advertising creative production. By 2023, the company further established the "AI Operations Office" to oversee the construction of an enterprise-level AI application framework and governance system at the organizational level. This clearly delineated developmental trajectory demonstrates CyberAgent's strategic positioning of AI as a core organizational asset rather than merely a technological tool.

Core Deployed Products and Tool Ecosystem

In terms of specific product deployment, CyberAgent adopted a dual-core tool strategy. ChatGPT Enterprise serves as a general-purpose AI assistant, primarily addressing daily office scenarios including research analysis, content creation, and information organization. Codex functions as a professional-grade programming assistant, covering specialized development workflows such as code review, design discussions, documentation, and development planning. This clearly differentiated tool configuration strategy not only satisfies the diverse business needs of the enterprise but also ensures deep application value in specialized scenarios.

Central Theme: Voluntary Adoption and Culture-Driven AI Integration

The most remarkable aspect of the CyberAgent case lies in its distinctive approach characterized by a "non-mandatory, voluntary adoption" strategy. Without implementing any compulsory usage policies, ChatGPT Enterprise achieved a remarkable 93% monthly active user rate, with usage spanning virtually all departments and over 100 employees participating in more than ten training sessions. This achievement subverts the conventional wisdom that "mandatory enforcement is necessary to ensure adoption rates" in traditional enterprise software deployment, revealing instead the possibilities that emerge when AI achieves deep organizational penetration through cultural construction and knowledge sharing.

In-Depth Analysis of Application Scenarios and Effectiveness Assessment

Multi-Scenario Application Practices of ChatGPT Enterprise

Within daily office operations, the application of ChatGPT Enterprise exhibits remarkable breadth and depth. Research analysts leverage it for rapid market intelligence consolidation and competitive analysis. Content operations teams utilize it for copywriting and creative brainstorming. Product managers employ it for structured documentation of requirements and efficient meeting minutes generation. Crucially, ChatGPT Enterprise does not simply replace human work; instead, it assumes the role of a "thinking partner," helping employees gain multi-dimensional reference information in complex decision-making scenarios.

In terms of information security, CyberAgent fully leveraged the enterprise-grade security capabilities of ChatGPT Enterprise, including account management, usage visibility, and access control. The company established a comprehensive internal guideline system that clearly delineates acceptable information types for AI tool input while implementing strict protection for confidential data. This security governance framework achieves an effective balance between AI application scalability and data protection.

Deep Integration of Codex in Development Workflows

The introduction of Codex brought significant transformation to CyberAgent's development workflow. In design review processes, Codex can comprehensively evaluate and stress-test design proposals from multiple perspectives, helping teams achieve more thorough consensus before implementation and significantly reducing rework caused by design flaws. Developer Hidekazu Hora remarked: "Codex functions like a reliable partner, supporting the entire process from discussing implementation approaches to execution, effectively enhancing development speed."

In the code review dimension, Codex not only generates improvement suggestions but also assists teams in selecting optimal options among multiple alternatives. Notably, Codex's value extends beyond mere coding speed improvement to systematic enhancement of development quality. As Sou Yoshihara, a senior Codex power user from the AI Business Division, evaluated: "Compared with other programming models, Codex gives the impression of producing higher-quality proposals. It is not merely a tool but rather a methodology for optimizing the overall development process."

Signature Project Cases: Kiwami Prediction AI and WormEscape

The Kiwami Prediction AI project deeply applied Codex's MCP (Model Context Protocol) capabilities during its design and implementation planning phases, achieving high-integration AI capability with the professional development environment through the Cursor editor. This case demonstrates how AI Agent capabilities can be seamlessly embedded within existing professional development toolchains.

The development cycle for the WormEscape game was completed for a soft launch in approximately one month, with Codex playing a pivotal role. This case powerfully validates AI Agent's practical value in accelerating product development cycles while demonstrating that AI can effectively help developers rapidly overcome knowledge barriers even in areas where they lack prior experience.

Utility Analysis and Value Assessment

Dual-Dimensional Examination of Quantitative Metrics and Qualitative Benefits

From a quantitative perspective, the 93% monthly active user rate, participation exceeding 100 employees per training session across more than ten sessions, and usage coverage spanning virtually all departments—these metrics fully validate the high penetration and acceptance of AI tools within CyberAgent. However, what deserves greater attention are the driving factors and sustainability mechanisms behind this success.

From a qualitative dimension, CyberAgent's AI application achieves multi-layered value: enhanced decision quality—through multi-perspective analysis supporting more comprehensive judgment; improved collaboration efficiency—the application of Codex in design reviews significantly reduced internal communication costs and rework frequency; strengthened knowledge transfer—AI tools emerged as effective supplementary means for newcomers to rapidly familiarize themselves with business and technology; unleashed innovation capacity—employees liberated from repetitive tasks channeled more energy into creative endeavors.

The Success Logic Behind the Non-Mandatory Strategy

CyberAgent's choice to forgo mandatory adoption policies achieved high penetration rates through the following mechanisms:

Knowledge sharing mechanisms constitute the core driving force. Internal promotion of effective prompts and successful application cases created a virtuous knowledge dissemination network. Rather than being compelled to use AI, employees proactively learned and experimented after witnessing high-value applications by colleagues. This bottom-up diffusion model possesses stronger sustainability and deeper penetration than top-down administrative mandates.

Visibility-based incentives likewise played a significant role. The company established an internal usage ranking system; while data was not used for performance evaluation, it provided employees with benchmarks for self-reference and target pursuit. This transparent feedback mechanism satisfied employees' cognitive needs for self-improvement while avoiding resistance stemming from coercion.

Automated follow-ups ensured implementation continuity. For employees who had not used the tools for extended periods, the system automatically sent reminders via Slack, though these follow-ups represented gentle guidance rather than mandatory requirements. This design respected employees' learning pace while ensuring sustained tool promotion.

Tiered training systems addressed differentiated needs. Training courses spanning from beginner to advanced levels covered employees of varying roles and skill levels, ensuring everyone could find a suitable learning path.

The Art of Balancing Security and Scalability

In advancing AI applications, CyberAgent fully recognized the prerequisite importance of security governance. Through establishing clear internal guidelines, strict account management systems, and usage visibility functions, the company effectively controlled information security risks while expanding AI application scope. As Ken Takao, Manager of the Data Technology Department, summarized: "With enterprise features such as account management and visibility into usage, ChatGPT Enterprise made it possible to support business use of a wide range of information, excluding confidential data. As a result, the scope of AI use across the company has expanded, and many employees now integrate AI into their daily work."

Inspirational Significance and the Elevation of AI Intelligence Applications

Universal Lessons for the Industry

CyberAgent's practices provide invaluable reference frameworks for enterprise-level AI Agent deployment. First and foremost, the priority of cultural construction should proceed in parallel with technology deployment. The achievement of a 93% active user rate reflects, on the surface, the success of tools, but at a deeper level, represents a triumph of organizational culture. When employees perceive AI as a partner enhancing their capabilities rather than a surveillance mechanism or replacement threat, voluntary adoption becomes the natural outcome.

Secondly, gradual expansion outperforms radical replacement. CyberAgent did not attempt to replace all work with AI in a single stride; instead, it progressively expanded AI application boundaries through continuous scenario discovery and successful case sharing. This strategy reduced transformation resistance, cultivated employees' AI literacy, and created conditions for subsequently deeper integration.

Thirdly, the value positioning of tools determines the depth of application. Positioning AI as a "quality judgment improvement tool" rather than a mere "speed enhancement tool" elevated Codex's application value beyond simple efficiency calculations, extending into higher dimensions such as decision quality, workflow optimization, and professional capability enhancement.

Industry Trend Insights on AI Agent Development

The CyberAgent case reflects several significant trends in the AI Agent field. From the technology integration dimension, AI agents are evolving from independent tools toward deeply embedded workflow components. The integration of Codex with Cursor through the MCP protocol demonstrates how AI capability can be seamlessly connected with professional development environments to unlock greater value.

From the role positioning dimension, AI agents are transitioning from "executors" to "collaborative partners." Employee feedback consistently emphasized AI's auxiliary value in discussion, review, and decision-making processes requiring human judgment, rather than merely replacement functions at the execution level.

From the governance model dimension, enterprise AI applications are forming a三位一体 (three-in-one) advancement paradigm of "security first, value-driven, culture-supported." Pure technology deployment cannot guarantee success; radical promotion lacking security frameworks carries substantial risks; and strategies lacking cultural support struggle to sustain.

Prospects for Intelligent Applications Toward the Future

CyberAgent regards AI as a pivotal technology that may become part of the next-generation internet industry standard. This judgment carries profound strategic insight. When AI capabilities become part of the work infrastructure, enterprise competitive advantages will no longer derive merely from "whether AI is used," but rather from "how AI is deeply integrated to unlock unique value."

For enterprises planning AI Agent deployment, the CyberAgent case provides a clear success pathway: establish a forward-looking AI strategic framework (such as the creation of an AI Operations Office); construct a comprehensive security governance system (application of enterprise-grade security features and establishment of internal guidelines); design culture-driven promotion mechanisms (knowledge sharing, voluntary adoption, tiered training); pursue deep integration rather than superficial application (embed AI into core workflows to enhance decision quality and development quality).

Conclusion

The CyberAgent AI Agent enterprise-level deployment case serves as a profound textbook on successfully transforming cutting-edge AI technology into organizational productivity. Behind its 93% monthly active user rate lies the power of culture rather than the pressure of coercion. The quality improvements brought by Codex reflect deep practice of human-machine collaboration philosophy rather than simple tool replacement logic.

The core value of this case lies in revealing the success equation for enterprise AI Agent deployment: advanced technological tools + comprehensive security governance + voluntarily-driven cultural mechanisms = sustainable deep application. As AI Agent technology continues to evolve, CyberAgent's experience reminds us that the decisive factor in technological success often lies not in the technology itself but in the depth of integration between technology, organization, and culture.

When Engineers at Anthropic Learn to Work with Claude

January 16, 2026

— A narrative and analytical review of How AI Is Transforming Work at Anthropic, focusing on personal efficiency, capability expansion, learning evolution, and professional identity in the AI era.

In November 2025, Anthropic released its research report How AI Is Transforming Work at Anthropic. After six months of study, the company did something unusual: it turned its own engineers into research subjects.

Across 132 engineers, 53 in-depth interviews, and more than 200,000 Claude Code sessions, the study aimed to answer a single fundamental question:

How does AI reshape an individual’s work? Does it make us stronger—or more uncertain?

The findings were both candid and full of tension:

Roughly 60% of engineering tasks now involve Claude, nearly double from the previous year;
Engineers self-reported an average productivity gain of 50%;
27% of AI-assisted tasks represented “net-new work” that would not have been attempted otherwise;
Many also expressed concerns about long-term skill degradation and the erosion of professional identity.

This article distills Anthropic’s insights through four narrative-driven “personal stories,” revealing what these shifts mean for knowledge workers in an AI-transformed workplace.

Efficiency Upgrades: When Time Is Reallocated, People Rediscover What Truly Matters

Story: From “Defusing Bombs” to Finishing a Full Day’s Work by Noon

Marcus, a backend engineer at Anthropic, maintained a legacy system weighed down by years of technical debt. Documentation was sparse, function chains were tangled, and even minor modifications felt risky.

Previously, debugging felt like bomb disposal:

checking logs repeatedly
tracing convoluted call chains
guessing root causes
trial, rollback, retry

One day, he fed the exception stack and key code segments into Claude.

Claude mapped the call chain, identified three likely causes, and proposed a “minimum-effort fix path.” Marcus’s job shifted to:

selecting the most plausible route,
asking Claude to generate refactoring steps and test scaffolds,
adjusting only the critical logic.

He finished by noon. The remaining hours went into discussing new product trade-offs—something he rarely had bandwidth for before.

Insight: Efficiency isn’t about “doing the same task faster,” but about “freeing attention for higher-value work.”

Anthropic’s data shows:

Debugging and code comprehension are the most frequent Claude use cases;
Engineers saved “a little time per task,” but total output expanded dramatically.

Two mechanisms drive this:

AI absorbs repeatable, easily verifiable, low-friction tasks, lowering the psychological cost of getting started;
Humans can redirect time toward analysis, decision-making, system design, and trade-off reasoning—where actual value is created.

This is not linear acceleration; it is qualitative reallocation.

Personal Takeaway: If you treat AI as a code generator, you’re using only 10% of its value.

What to delegate:

log diagnosis
structural rewrites
boilerplate implementation
test scaffolding
documentation framing

Where to invest your attention:

defining the problem
architectural trade-offs
code review
cross-team alignment
identifying the critical path

What you choose to work on—not how fast you type—is where your value lies.

Capability Expansion: When Cross-Stack Work Stops Being Intimidating

Story: A Security Engineer Builds the First Dashboard of Her Life

Lisa, a member of the security team, excelled at threat modeling and code audits—but had almost no front-end experience.

The team needed a real-time risk dashboard. Normally this meant:

queuing for front-end bandwidth,
waiting days or weeks,
iterating on a minimal prototype.

This time, she fed API response data into Claude and asked:

“Generate a simple HTML + JS interface with filters and basic visualization.”

Within seconds, Claude produced a working dashboard—charts, filters, and interactions included.
Lisa polished the styling and shipped it the same day.

For the first time, she felt she could carry a full problem from end to end.

Insight: AI turns “I can’t do this” into “I can try,” and “try” into “I can deliver.”

One of the clearest conclusions from Anthropic’s report:

Everyone is becoming more full-stack.

Evidence:

Security teams navigate unfamiliar codebases with AI;
Researchers create interactive data visualizations;
Backend engineers perform lightweight data analysis;
Non-engineers write small automation scripts.

This doesn’t eliminate roles—it shortens the path from idea to MVP, deepens end-to-end system understanding, and raises the baseline capability of every contributor.

Personal Takeaway: The most valuable skill isn’t a specific tech stack—it's how quickly AI amplifies your ability to cross domains.

Practice:

Use AI for one “boundary task” you’re not familiar with (front end, analytics, DevOps scripts).
Evaluate the reliability of the output.
Transfer the gained understanding back into your primary role.

In the AI era, your identity is no longer “backend/front-end/security/data,”
but:

Can you independently close the loop on a problem?

Learning Evolution: AI Accelerates Doing, but Can Erode Understanding

Story: The New Engineer Who “Learns Faster but Understands Less”

Alex, a new hire, needed to understand a large service mesh.
With Claude’s guidance, he wrote seemingly reasonable code within a week.

Three months later, he realized:

he knew how to write code, but not why it worked;
Claude understood the system better than he did;
he could run services, but couldn’t explain design rationale or inter-service communication patterns.

This was the “supervision paradox” many engineers described:

To use AI well, you must be capable of supervising it—
but relying on AI too heavily weakens the very ability required for supervision.

Insight: AI accelerates procedural learning but dilutes conceptual depth.

Two speeds of learning emerge:

Procedural learning (fast): AI provides steps and templates.
Conceptual learning (slow): Requires structural comprehension, trade-off reasoning, and system thinking.

AI creates the illusion of mastery before true understanding forms.

**Personal Takeaway: Growth comes from dialogue with AI, not delegation to AI.**

To counterbalance the paradox:

Write a first draft yourself before asking AI to refine it.
Maintain “no-AI zones” for foundational practice.
Use AI as a teacher:
- ask for trade-off explanations,
- compare alternative architectures,
- request detailed code review logic,
- force yourself to articulate “why this design works.”

AI speeds you up, but only you can build the mental models.

Professional Identity: Between Excitement and Anxiety

Story: Some Feel Like “AI Team Leads”—Others Feel Like They No Longer Write Code

Reactions varied widely:

Some engineers said:

“It feels like managing a small AI engineering team. My output has doubled.”
Others lamented:

“I enjoy writing code. Now my work feels like stitching together AI outputs. I’m not sure who I am anymore.”

A deeper worry surfaced:

“If AI keeps improving, what remains uniquely mine?”

Anthropic doesn’t offer simple reassurance—but reveals a clear shift:

Professional identity is moving from craft execution to system orchestration.

**Insight: The locus of human value is shifting from doing tasks to directing how tasks get done.**

AI already handles:

coding
debugging
test generation
documentation scaffolding

But it cannot replace:

contextual judgment across team, product, and organization
long-term architectural reasoning
multi-stakeholder coordination
communication, persuasion, and explanation

These human strengths become the new core competencies.

Personal Takeaway: Your value isn’t “how much you code,” but “how well you enable code to be produced.”

Ask yourself:

Do I know how to orchestrate AI effectively in workflows and teams?
Can I articulate why a design choice is better than alternatives?
Am I shifting from executor to designer, reviewer, or coordinator?

If yes, your career is already evolving upward.

An Anthropic-Style Personal Growth Roadmap

Putting the four stories together reveals an “AI-era personal evolution model”:

1. Efficiency Upgrade: Reclaim attention from low-value zones

AI handles: repetitive, verifiable, mechanical tasks
You focus on: reasoning, trade-offs, systemic thinking

2. Capability Expansion: Cross-stack and cross-domain agility becomes the norm

AI lowers technical barriers
You turn lower barriers into higher ownership

3. Learning Evolution: Treat AI as a sparring partner, not a shortcut

AI accelerates doing
You consolidate understanding
Contrast strengthens judgment

4. Professional Identity Shift: Move toward orchestration and supervision

AI executes
You design, interpret, align, and guide

One-Sentence Summary

Anthropic shows how individuals become stronger—not by coding faster, but by redefining their relationship with AI and elevating themselves into orchestrators of human-machine collaboration.

Intelligent Evolution of Individuals and Organizations: How Harvey Is Bringing AI Productivity to Ground in the Legal Industry

January 11, 2026

Over the past two years, discussions around generative AI have often focused on model capability improvements. Yet the real force reshaping individuals and organizations comes from products that embed AI deeply into professional workflows. Harvey is one of the most representative examples of this trend.

As an AI startup dedicated to legal workflows, Harvey reached a valuation of 8 billion USD in 2025. Behind this figure lies not only capital market enthusiasm, but also a profound shift in how AI is reshaping individual career development, professional division of labor, and organizational modes of production.

This article takes Harvey as a case study to distill the underlying lessons of intelligent productivity, offering practical reference to individuals and organizations seeking to leverage AI to enhance capabilities and drive organizational transformation.

The Rise of Vertical AI: From “Tool” to “Operating System”

Harvey’s rapid growth sends a very clear signal.

Total financing in the year: 760 million USD
Latest round: 160 million USD, led by a16z
Annual recurring revenue (ARR): 150 million USD, doubling year-on-year
User adoption: used by around 50% of Am Law 100 firms in the United States

These numbers are more than just signs of investor enthusiasm; they indicate that vertical AI is beginning to create structural value in real industries.

The evolution of generative AI roughly经历了三个阶段：

Phase 1: Public demonstrations of general-purpose model capabilities
Phase 2: AI-driven workflow redesign for specific professional scenarios
Phase 3 (where Harvey now operates): becoming an industry operating system for work

In other words, Harvey is not simply a “legal GPT”. It is a complete production system that combines:

Model capabilities + compliance and governance + workflow orchestration + secure data environments

For individual careers and organizational structures, this marks a fundamentally new kind of signal:

AI is no longer just an assistive tool; it is a powerful engine for restructuring professional division of labor.

How AI Elevates Professionals: From “Tool Users” to “Designers of Automated Workchains”

Harvey’s stance is explicit: “AI will not replace lawyers; it replaces the heavy lifting in their work.”
The point here is not comfort messaging, but a genuine shift in the logic of work division.

A lawyer’s workchain is highly structured:
Research → Reading → Reasoning → Drafting → Reviewing → Delivering → Client communication

With AI in the loop, 60–80% of this chain can be standardized, automated, and reused at scale.

How It Enhances Individual Professional Capability

Task Completion Speed Increases Dramatically
Time-consuming tasks such as drafting documents, compliance reviews, and case law research are handled by AI, freeing lawyers to focus on strategy, litigation preparation, and client relations.
Cognitive Boundaries Are Expanded
AI functions like an “infinitely extendable external brain”, enabling professionals to construct deeper and broader understanding frameworks in far less time.
Capability Becomes More Transferable Across Domains
Unlike traditional division of labor, where experience is locked in specific roles or firms, AI-driven workflows help individuals codify methods and patterns, making it easier to transfer and scale their expertise across domains and scenarios.

In this sense, the most valuable professionals of the future are not just those who “possess knowledge”, but those who master AI-powered workflows.

Organizational Intelligent Evolution: From Process Optimization to Production Model Transformation

Harvey’s emergence marks the first production-model-level transformation in the legal sector in roughly three decades.
The lessons here extend far beyond law and are highly relevant for all types of organizations.

1. AI Is Not Just About Efficiency — It Redesigns How People Collaborate

Harvey’s new product — a shared virtual legal workspace — enables in-house teams and law firms to collaborate securely, with encrypted isolation preventing leakage of sensitive data.

At its core, this represents a new kind of organizational design:

Work is no longer constrained by physical location
Information flows are no longer dependent on manual handoffs
Legal opinions, contracts, and case law become reusable, orchestratable building blocks
Collaboration becomes a real-time, cross-team, cross-organization network

These shifts imply a redefinition of organizational boundaries and collaboration patterns.

2. AI Is Turning “Unstructured Problems” in Complex Industries Into Structured Ones

The legal profession has long been seen as highly dependent on expertise and judgment, and therefore difficult to standardize. Harvey demonstrates that:

Data can be structured
Reasoning chains can be modeled
Documents can be generated and validated automatically
Risk and compliance can be monitored in real time by systems

Complex industries are not “immune” to AI transformation — they simply require AI product teams that truly understand the domain.

The same pattern will quickly replicate in consulting, investment research, healthcare, insurance, audit, tax, and beyond.

3. Organizations Will Shift From “Labor-Intensive” to “Intelligence-Intensive”

In an AI-driven environment, the ceiling of organizational capability will depend less on how many people are hired, and more on:

How many workflows are genuinely AI-automated
Whether data can be understood by models and turned into executable outputs
Whether each person can leverage AI to take on more decision-making and creative tasks

In short, organizational competitiveness will increasingly hinge on the depth and breadth of intelligentization, rather than headcount.

The True Value of Vertical AI SaaS: From Wrapping Models to Encapsulating Industry Knowledge

Harvey’s moat does not come from having “a better model”. Its defensibility rests on three dimensions:

1. Deep Workflow Integration

From case research to contract review, Harvey is embedded end-to-end in legal workflows.
This is not “automating isolated tasks”, but connecting the entire chain.

2. Compliance by Design

Security isolation, access control, compliance logs, and full traceability are built into the product.
In legal work, these are not optional extras — they are core features.

3. Accumulation and Transfer of Structured Industry Knowledge

Harvey is not merely a frontend wrapper around GPT. It has built:

A legal knowledge graph
Large-scale embeddings of case law
Structured document templates
A domain-specific workflow orchestration engine

This means its competitive moat lies in long-term accumulation of structured industry assets, not in any single model.

Such a product cannot be easily replaced by simply swapping in another foundation model. This is precisely why top-tier investors are willing to back Harvey at such a scale.

Lessons for Individuals, Organizations, and Industries: AI as a New Platform for Capability

Harvey’s story offers three key takeaways for broader industries and for individual growth.

Insight 1: The Core Competency of Professionals Is Shifting From “Owning Knowledge” to “Owning Intelligent Productivity”

In the next 3–5 years, the rarest and most valuable talent across industries will be those who can:

Harness AI, design AI-powered workflows, and use AI to amplify their impact.

Every professional should be asking:

Can I let AI participate in 50–70% of my daily work?
Can I structure my experience and methods, then extend them via AI?
Can I become a compounding node for AI adoption in my organization?

Mastering AI is no longer a mere technical skill; it is a career leverage point.

Insight 2: Organizational Intelligentization Depends Less on the Model, and More on Whether Core Workflows Can Be Rebuilt

The central question every organization must confront is:

Do our core workflows already provide the structural space needed for AI to create value?

To reach that point, organizations need to build:

Data structures that can be understood and acted upon by models
Business processes that can be orchestrated rather than hard-coded
Decision chains where AI can participate as an agent rather than as a passive tool
Automated systems for risk and compliance monitoring

The organizations that ultimately win will be those that can design robust human–AI collaboration chains.

Insight 3: The Vertical AI Era Has Begun — Winners Will Be Those Who Understand Their Industry in Depth

Harvey’s success is not primarily about technology. It is about:

Deep understanding of the legal domain
Deep integration into real legal workflows
Structural reengineering of processes
Gradual evolution into industry infrastructure

This is likely to be the dominant entrepreneurial pattern over the next decade.

Whether the arena is law, climate, ESG, finance, audit, supply chain, or manufacturing, new “operating systems for industries” will continue to emerge.

Conclusion: AI Is Not Replacement, but Extension; Not Assistance, but Reinvention

Harvey points to a clear trajectory:

AI does not primarily eliminate roles; it upgrades them.
It does not merely improve efficiency; it reshapes production models.
It does not only optimize processes; it rebuilds organizational capabilities.

For individuals, AI is a new amplifier of personal capability.
For organizations, AI is a new operating system for work.
For industries, AI is becoming new infrastructure.

The era of vertical AI has genuinely begun.
The real opportunities belong to those willing to redefine how work is done and to actively build intelligent organizational capabilities around AI.

From System Records to Agent Records: Workday’s Enterprise AI Transformation Paradigm—A Future of Human–Digital Agent Coexistence

October 13, 2025

Based on a McKinsey Inside the Strategy Room interview with Workday CEO Carl Eschenbach (August 21, 2025), combined with Workday official materials and third-party analyses, this study focuses on enterprise transformation driven by agentic AI. Workday’s practical experience in human–machine collaborative intelligence offers valuable insights.

In enterprise AI transformation, two extremes must be avoided: first, treating AI as a “universal cost-cutting tool,” falling into the illusion of replacing everything while neglecting business quality, risk, and experience; second, refusing to experiment due to uncertainty, thereby missing opportunities to elevate efficiency and value.

The proper approach positions AI as a “productivity-enhancing digital colleague” under a governance and measurement framework, aiming for measurable productivity gains and new value creation. By starting with small pilots and iterative scaling, cost reduction, efficiency enhancement, and innovation can be progressively unified.

Overview

Workday’s AI strategy follows a “human–agent coexistence” paradigm. Using consistent data from HR and finance systems of record (SOR) and underpinned by governance, the company introduces an “Agent System of Record (ASR)” to centrally manage agent registration, permissions, costs, and performance—enabling a productivity leap from tool to role-based agent.

Key Principles and Concepts

Coexistence, Not Replacement: AI’s power comes from being “agentic”—technology working for you. Workday clearly positions AI for peaceful human–agent coexistence.
Domain Data and Business Context Define the Ceiling: The CEO emphasizes that data quality and domain context, especially in HR and finance, are foundational. Workday serves over 10,000 enterprises, accumulating structured processes and data assets across clients.
Three-System Perspective: HR, finance, and customer SORs form the enterprise AI foundation. Workday focuses on the first two and collaborates with the broader ecosystem (e.g., Salesforce).
Speed and Culture as Multipliers: Treating “speed” as a strategic asset and cultivating a growth-oriented culture through service-oriented leadership that “enables others.”

Practice and Governance (Workday Approach)

ASR Platform Governance: Unified directories and observability for centralized control of in-house and third-party agents; role and permission management, registration and compliance tracking, cost budgeting and ROI monitoring, real-time activity and strategy execution, and agent orchestration/interconnection via A2A/MCP protocols (Agent Gateway). Digital colleagues in HaxiTAG Bot Factory provide similar functional benefits in enterprise scenarios.
Role-Based (Multi-Skill) Agents: Upgrade from task-based to configurable “role” agents, covering high-value processes such as recruiting, talent mobility, payroll, contracts, financial audit, and policy compliance.
Responsible AI System: Appoint a Chief Responsible AI Officer and employ ISO/IEC 42001 and NIST AI RMF for independent validation and verification, forming a governance loop for bias, security, explainability, and appeals.
Organizational Enablement: Systematic AI training for 20,000+ employees to drive full human–agent collaboration.

Value Proposition and Business Implications

From “Application-Centric” to “Role-Agent-Centric” Experience: Users no longer “click apps” but collaborate with context-aware role agents, requiring rethinking of traditional UI and workflow orchestration.
Measurable Digital Workforce TCO/ROI: ASR treats agents as “digital employees,” integrating budget, cost, performance, and compliance into a single ledger, facilitating CFO/CHRO/CAIO governance and investment decisions.
Ecosystem and Interoperability: Agent Gateway connects external agents (partners or client-built), mitigating “agent sprawl” and shadow IT risks.

Methodology: A Reusable Enterprise Deployment Framework

Objective Function: Maximize productivity, minimize compliance/risk, and enhance employee experience; define clear boundaries for tasks agents can independently perform.
Priority Scenarios: Select high-frequency, highly regulated, and clean-data HR/finance processes (e.g., payroll verification, policy responses, compliance audits, contract obligation extraction) as MVPs.
ASR Capability Blueprint:
- Directory: Agent registration, profiles (skills/capabilities), tracking, explainability;
- Identity & Permissions: Least privilege, cross-system data access control;
- Policy & Compliance: Policy engine, action audits, appeals, accountability;
- Economics: Budgeting, A/B and performance dashboards, task/time/result accounting;
- Connectivity: Agent Gateway, A2A/MCP protocol orchestration.
“Onboard Agents Like Humans”: Implement lifecycle management and RACI assignment for “hire–trial–performance–promotion–offboarding” to prevent over-authorization or improper execution.
Responsible AI Governance: Align with ISO 42001 and NIST AI RMF; establish processes and metrics (risk registry, bias testing, explainability thresholds, red teaming, SLA for appeals), and regularly disclose internally and externally.
Organization and Culture: Embed “speed” in OKRs/performance metrics, emphasize leadership in “serving others/enabling teams,” and establish CAIO/RAI committees with frontline coaching mechanisms.

Industry Insight: Instead of full-scale rollout, adopt a four-piece “role–permission–metric–governance” loop, gradually delegating authority to create explainable autonomy.

Assessment and Commentary

Workday unifies humans and agents within existing HR/finance SORs and governance, balancing compliance with practical deployment density, shortening the path from pilot to scale. Constraints and risks include:

Ecosystem Lock-In: ASR strongly binds to Workday data and processes; open protocols and Marketplace can mitigate this.
Cross-System Consistency: Agents spanning ERP/CRM/security domains require end-to-end permission and audit linkage to avoid “shadow agents.”
Measurement Complexity: Agent value must be assessed by both process and outcome (time saved ≠ business result).

Sources: McKinsey interview with Workday CEO on “coexistence, data quality, three-system perspective, speed and leadership, RAI and training”; Workday official pages/news on ASR, Agent Gateway, role agents, ROI, and Responsible AI; HFS, Josh Bersin, and other industry analyses on “agent sprawl/governance.”

Morgan Stanley’s DevGen.AI: Reshaping Enterprise Legacy System Modernization Through Generative AI

September 09, 2025

As enterprises increasingly grapple with the pressing challenge of modernizing legacy software systems, Morgan Stanley has unveiled DevGen.AI—an internally developed generative AI tool that sets a new benchmark for enterprise-grade modernization strategies. Built upon OpenAI’s GPT models, DevGen.AI is designed to tackle the long-standing issue of outdated systems—particularly those written in languages like COBOL—that are difficult to maintain, adapt, or scale within financial institutions.

The Innovation: A Semantic Intermediate Layer

DevGen.AI’s most distinctive innovation lies in its use of an “intermediate language” approach. Rather than directly converting legacy code into modern programming languages, it first translates source code into structured, human-readable English specifications. Developers can then use these specs to rewrite the system in modern languages. This human-in-the-loop paradigm—AI-assisted specification generation followed by manual code reconstruction—offers superior adaptability and contextual accuracy for the modernization of complex, deeply embedded enterprise systems.

By 2025, DevGen.AI has analyzed over 9 million lines of legacy code, saving developers more than 280,000 working hours. This not only reduces reliance on scarce COBOL expertise but also provides a structured pathway for large-scale software asset refactoring across the firm.

Application Scenarios and Business Value at Morgan Stanley

DevGen.AI has been deployed across three core domains:

1. Code Modernization & Migration

DevGen.AI accelerates the transformation of decades-old mainframe systems by translating legacy code into standardized technical documentation. This enables faster and more accurate refactoring into modern languages such as Java or Python, significantly shortening technology upgrade cycles.

2. Compliance & Audit Support

Operating in a heavily regulated environment, financial institutions must maintain rigorous transparency. DevGen.AI facilitates code traceability by extracting and describing code fragments tied to specific business logic, helping streamline both internal audits and external regulatory responses.

3. Assisted Code Generation

While its generated modern code is not yet fully optimized for production-scale complexity, DevGen.AI can autonomously convert small to mid-sized modules. This provides substantial savings on initial development efforts and lowers the barrier to entry for modernization.

A key reason for Morgan Stanley’s choice to build a proprietary AI tool is the ability to fine-tune models based on domain-specific semantics and proprietary codebases. This avoids the semantic drift and context misalignment often seen with general-purpose LLMs in enterprise environments.

Strategic Insights from an AI Engineering Milestone

DevGen.AI exemplifies a systemic response to technical debt in the AI era, offering a replicable roadmap for large enterprises. Beyond showcasing generative AI’s real-world potential in complex engineering tasks, the project highlights three transformative industry trends:

1. Legacy System Integration Is the Gateway to Industrial AI Adoption

Enterprise transformation efforts are often constrained by the inertia of legacy infrastructure. DevGen.AI demonstrates that AI can move beyond chatbot interfaces or isolated coding tasks, embedding itself at the heart of IT infrastructure transformation.

2. Semantic Intermediation Is Critical for Quality and Control

By shifting the translation paradigm from “code-to-code” to “code-to-spec,” DevGen.AI introduces a bilingual collaboration model between AI and humans. This not only enhances output fidelity but also significantly improves developer control, comprehension, and confidence.

3. Organizational Modernization Amplifies AI ROI

Mike Pizzi, Morgan Stanley’s Head of Technology, notes that AI amplifies existing capabilities—it is not a substitute for foundational architecture. Therefore, the success of AI initiatives hinges not on the models themselves, but on the presence of a standardized, modular, and scalable technical infrastructure.

From Intelligent Tools to Intelligent Architecture

DevGen.AI proves that the core enterprise advantage in the AI era lies not in whether AI is adopted, but in how AI is integrated into the technology evolution lifecycle. AI is no longer a peripheral assistant; it is becoming the central engine powering IT transformation.

Through DevGen.AI, Morgan Stanley has not only addressed legacy technical debt but has also pioneered a scalable, replicable, and sustainable modernization framework. This breakthrough sets a precedent for AI-driven transformation in highly regulated, high-complexity industries such as finance. Ultimately, the value of enterprise AI does not reside in model size or novelty—but in its strategic ability to drive structural modernization.

The Strategic Shift of Generative AI in the Enterprise: From Adoption Surge to Systemic Evolution

August 01, 2025

Bain & Company’s report, “Despite Barriers, the Adoption of Generative AI Reaches an All-Time High”, provides an authoritative and structured exploration of the strategic significance, systemic challenges, and capability-building imperatives of generative AI (GenAI) in enterprise services. It offers valuable insights for senior executives and technical leaders seeking to understand the business impact and organizational implications of GenAI deployment.

Generative AI at Scale: A Technological Leap Triggering Organizational Paradigm Shifts

According to Bain’s 2025 survey, 95% of U.S. enterprises have adopted generative AI, with production use cases increasing by 101% year-over-year. This leap signals not only technological maturity but a foundational shift in enterprise operating models—GenAI is no longer a peripheral innovation but a core driver reshaping workflows, customer engagement, and product development.

The IT function has emerged as the fastest adopter, integrating GenAI into modules such as code generation, knowledge retrieval, and system operations—demonstrating the technology’s natural alignment with knowledge-intensive tasks. Initially deployed to enhance operational efficiency and reduce costs, GenAI is now evolving from a productivity enhancer into a value creation engine as enterprises deepen its application.

Strategic Prioritization: Evolving Enterprise Mindsets and Readiness Gaps

Notably, the share of companies prioritizing AI as a strategic initiative has risen to 15% within a year, and 50% now have a defined implementation roadmap. This trend indicates a shift among leading firms from a narrow focus on deployment to building comprehensive AI governance frameworks—encompassing platform architecture, talent models, data assets, and process redesign.

However, the report also reveals a significant bifurcation: half of all companies still lack a clear strategy. This reflects an emerging “capability polarization” in the market. Front-runners are institutionalizing GenAI through standardized workflows, mature governance, and deep vendor partnerships, while others remain stuck in fragmented pilots without coherent organizational frameworks.

Realizing Value: A Reinforcing Feedback Loop of Performance and Confidence

Over 80% of reported use cases met or exceeded expectations, and nearly 60% of satisfied enterprises reported measurable business improvements—affirming the commercial viability of GenAI. These high-yield use cases—document generation, customer inquiry automation, internal search, reporting—share common traits: high knowledge structure, task repeatability, and stable context.

More importantly, this success has triggered a confidence flywheel: early wins → increased executive trust → expanded resource allocation → greater capabilities. Among organizations that have scaled GenAI, approximately 90% report target attainment or outperformance—highlighting the compounding marginal value of GenAI as it evolves from a tactical tool to a strategic platform.

Structural Challenges: Beyond Technical Hurdles to Organizational Complexity

Despite steep adoption curves, enterprises face three core, systemic constraints that must be addressed:

Data Security and Governance: As GenAI embeds itself deeper into critical systems, issues such as compliance, access control, and context integrity become paramount. Late-stage adopters are particularly focused on data lifecycle integrity and output accountability—underscoring the growing sensitivity to AI-related risk externalities.
Talent Gaps and Knowledge Asymmetries: 75% of companies report an inability to find internal expertise in critical functions. This is less about a shortage of AI engineers, and more about the lack of organizational infrastructure to integrate business users with AI systems—via interfaces, training, and process alignment.
Vendor Fragmentation and Ecosystem Fragility: With rapid evolution in AI infrastructure and models, long-term stability remains elusive. Concerns about vendor quality and model maintainability are surging among advanced adopters—reflecting increased strategic dependence on reliable ecosystem partners.

Reconstructing the Investment Rhythm: From Exploration Budgets to Operational Expenditures

Enterprise GenAI investment is entering a phase of structural normalization. Since early 2024, average annual AI budgets have reached $10 million—up 102% year-over-year. More significantly, 60% of GenAI projects are now funded through standard operating budgets, signaling a shift from experimental spending to institutionalized resource allocation.

This transition reflects a change in organizational perception: GenAI is no longer a one-off innovation initiative, but a core pillar within digital architecture, talent strategy, and process transformation. Enterprises are integrating GenAI into AI governance hubs and scenario-driven microservice deployments, emphasizing long-term, scalable orchestration.

Strategic Insight: GenAI as a Competitive Operating System of the Future

The central insight from Bain’s research is clear: generative AI is not just about technical deployment—it demands a fundamental redesign of organizational capabilities and cognitive infrastructure. Companies that sustainably unlock value from GenAI exhibit four shared traits:

Clear prioritization of high-value GenAI scenarios across the enterprise;
A cross-functional AI operations hub to align data, processes, models, and personnel;
A layered AI talent architecture—including prompt engineers, data governance experts, and domain modelers;
Integration of GenAI into core governance systems such as budgeting, KPIs, compliance, ethics, and knowledge management.

In the coming years, enterprise competition will no longer hinge on whether GenAI is adopted, but on how effectively organizations rewire their business models, restructure internal systems, and build defensible, sustainable AI capabilities. GenAI will become a benchmark for digital maturity—and a decisive differentiator in asymmetric competition.

Conclusion

Bain’s research offers a mirror reflecting how deeply generative AI is transforming the enterprise landscape. In this era of complex technological and organizational convergence, companies must look beyond tools and models. Strategic vision, systemic governance, and human-AI symbiosis are essential to unleashing the full multiplier effect of GenAI. Only with such a holistic approach can organizations seize the opportunity to lead in the next wave of digital transformation—and shape the future of business itself.

AI Automation: A Strategic Pathway to Enterprise Intelligence in the Era of Task Reconfiguration

July 13, 2025

With the rapid advancement of generative AI and task-level automation, the impact of AI on the labor market has gone far beyond the simplistic notion of "job replacement." It has entered a deeper paradigm of task reconfiguration and value redistribution. This transformation not only reshapes job design but also profoundly reconstructs organizational structures, capability boundaries, and competitive strategies. For enterprises seeking intelligent transformation and enhanced service and competitiveness, understanding and proactively embracing this change is no longer optional—it is a strategic imperative.

The "Dual Pathways" of AI Automation: Structural Transformation of Jobs and Skills

AI automation is reshaping workforce structures along two main pathways:

Routine Automation (e.g., customer service responses, schedule planning, data entry): By replacing predictable, rule-based tasks, automation significantly reduces labor demand and improves operational efficiency. A clear outcome is the decline in job quantity and the rise in skill thresholds. For instance, British Telecom’s plan to cut 40% of its workforce and Amazon’s robot fleet surpassing its human workforce exemplify enterprises adjusting the human-machine ratio to meet cost and service response imperatives.
Complex Task Automation (e.g., roles involving analysis, judgment, or interaction): Automation decomposes knowledge-intensive tasks into standardized, modular components, expanding employment access while lowering average wages. Job roles like telephone operators or rideshare drivers are emblematic of this "commoditization of skills." Research by MIT reveals that a one standard deviation drop in task specialization correlates with an 18% wage decrease—even as employment in such roles doubles, illustrating the tension between scaling and value compression.

For enterprises, this necessitates a shift from role-centric to task-centric job design, and a comprehensive recalibration of workforce value assessment and incentive systems.

Task Reconfiguration as the Engine of Organizational Intelligence: Not Replacement, but Reinvention

When implementing AI automation, businesses must discard the narrow view of “human replacement” and adopt a systems approach to task reengineering. The core question is not who will be replaced, but rather:

Which tasks can be automated?
Which tasks require human oversight?
Which tasks demand collaborative human-AI execution?

By clearly classifying task types and redistributing responsibilities accordingly, enterprises can evolve into truly human-machine complementary organizations. This facilitates the emergence of a barbell-shaped workforce structure: on one end, highly skilled "super-individuals" with AI mastery and problem-solving capabilities; on the other, low-barrier task performers organized via platform-based models (e.g., AI operators, data labelers, model validators).

Strategic Recommendations:

Accelerate automation of procedural roles to enhance service responsiveness and cost control.
Reconstruct complex roles through AI-augmented collaboration, freeing up human creativity and judgment.
Shift organizational design upstream, reshaping job archetypes and career development around “task reengineering + capability migration.”

Redistribution of Competitive Advantage: Platform and Infrastructure Players Reshape the Value Chain

AI automation is not just restructuring internal operations—it is redefining the industry value chain.

Platform enterprises (e.g., recruitment or remote service platforms) have inherent advantages in standardizing tasks and matching supply with demand, giving them control over resource allocation.
AI infrastructure providers (e.g., model developers, compute platforms) build strategic moats in algorithms, data, and ecosystems, exerting capability lock-in effects downstream.

To remain competitive, enterprises must actively embed themselves within the AI ecosystem, establishing an integrated “technology–business–talent” feedback loop. The future of competition lies not between individual companies, but among ecosystems.

Societal and Ethical Considerations: A New Dimension of Corporate Responsibility

AI automation exacerbates skill stratification and income inequality, particularly in low-skill labor markets, where “new structural unemployment” is emerging. Enterprises that benefit from AI efficiency gains must also fulfill corresponding responsibilities:

Support workforce skill transition through internal learning platforms and dual-capability development (“AI literacy + domain expertise”).
Participate in public governance by collaborating with governments and educational institutions to promote lifelong learning and career retraining systems.
Advance AI ethics governance to ensure fairness, transparency, and accountability in deployment, mitigating hidden risks such as algorithmic bias and data discrimination.

AI Is Not Destiny, but a Matter of Strategic Choice

As one industry mentor aptly stated, “AI is not fate—it is choice.” How a company defines which tasks are delegated to AI essentially determines its service model, organizational form, and value positioning. The future will not be defined by “AI replacing humans,” but rather by “humans redefining themselves through AI.”

Only by proactively adapting and continuously evolving can enterprises secure their strategic advantage in this era of intelligent reconfiguration.