Japanese version:
AIはなぜ経営成果に繋がらないのか 〜Agentic AI時代のオペレーティングモデル設計〜
*The views expressed in this article are based on the author’s personal experience and perspective and do not represent the official views or product direction of ServiceNow.
Introduction: Why Are Companies “Using AI” but Not Seeing Results?
Many companies are increasing their investments in AI. However, in most cases, these investments are not translating into business outcomes.
At the root of this issue is a structural problem: despite significant DX investments, strategy, investment, and execution remain fragmented, and companies have failed to build an operating model that continuously generates business value.
(Operating Model: the management structure that connects strategy to execution and enables a continuous cycle of decision-making, execution, and learning.)
The issue is not whether individual projects succeed or fail. Rather, it is the absence of a management structure that can immediately translate strategic changes into organizational execution.
This is where the concept of the Operating Model (OM) becomes critical.
According to McKinsey research, more than 78% of companies are already using generative AI in at least one business domain. At the same time, over 80% report that they are not seeing a material impact on revenue. This is the “Generative AI Paradox.”
The problem is not model accuracy or data volume. The core issue is that companies have not designed an operating model that connects strategy, investment, and execution with AI as a foundational assumption.
1. “Using AI” ≠ “Changing How Work Gets Done”
Today, generative AI is primarily used as a support tool—for drafting emails, summarization, or search. Decision-making, approval processes, and execution flows remain unchanged.
“Model accuracy is insufficient,” “data quality or quantity is lacking,” or “introducing new tools will solve the problem”—none of these address the root cause.
The real issue is that business processes are still designed around a human-centric operating model.
In the Agentic AI era, work and decision-making must be redesigned with execution entities that assume collaboration between humans and AI agents.
2. Why Previous AI Initiatives Failed to Deliver Business Impact
Figure 1: The Structural Gap Between AI Investment and Business Outcomes
The structural barriers preventing scale can be categorized into three layers.
| Overemphasis on “horizontal” use cases and stagnation of “vertical” ones | General-purpose tools are easy to deploy but tend to dilute impact, making contributions to revenue metrics unclear. Meanwhile, approximately 90% of domain-specific use cases remain stuck at the pilot stage and never reach full-scale deployment. |
| Limitations of “bolting AI onto” existing processes | When AI is positioned as a passive assistant and human-centric processes remain intact, productivity gains are typically limited to around 5–10%. |
| The “six barriers” to scale | A combination of insufficient strategy, talent shortages, technical limitations, organizational silos, data challenges, and cultural resistance interact to prevent AI initiatives from scaling. |
3. The Required Shift in Design Philosophy for the AI Era
Figure 2: Structural Barriers to Scaling AI
The solution is not a lack of technology, but a misalignment in design philosophy. Three fundamental shifts are required.
| From tool adoption to redesigning execution entities (humans × AI agents) | AI should be treated not as an assistive tool but as an “agent” capable of planning and completing tasks autonomously. This is realized through cross-functional workflow automation integrated with core systems. |
| From process improvement to operating model transformation | Rather than adding AI to existing processes, workflows must be rebuilt assuming AI autonomy. In call center examples, redesigning processes around AI agents has reduced resolution times by 60–90% and enabled autonomous resolution of 80% of Level 1 incidents. |
| From departmental optimization to CEO-led enterprise transformation | Organizations must move beyond fragmented experimentation and elevate AI adoption to a CEO-sponsored enterprise-wide strategic program. Key enablers include cross-functional transformation teams and adoption of an Agentic AI Mesh. |
4. Designing Management for AI Utilization (The Operating Model Perspective)
Figure 3: Three Required Design Shifts in the AI Era
An operating model is the structure that enables organizations to continuously execute decision‑making, execution, and learning as strategies evolve over time.
The success of AI adoption does not depend on individual use cases or tool selection, but on the readiness of four management foundations.
| People |
The success of AI investment equals the organization’s ability to execute transformation. People are management assets responsible for redesigning work and making decisions. |
| Governance | A new risk domain: how to govern autonomous software. Question: “What should be delegated to AI, and who is accountable?” |
| Technology | A shift from PoCs to an operational foundation that supports core business processes. Question: “Is this architecture designed to scale enterprise-wide?” |
| Data | Data is not an IT asset, but a management resource that drives operations. Question: “Has data become ‘actionable data’?” |
5. The Management Execution Platform for Implementing an Agentic AI Operating Model
Figure 4: The AI-Era Operating Model (Decision–Execution–Learning Loop)
To make AI an active execution entity, organizations need more than additional tools—they need a platform that runs execution.
ServiceNow overlays existing Systems of Record (SoR) and functions as a management execution platform that synchronizes strategy, investment, execution, and learning.
| Sense | Centralized AI visibility, governance, and usage management across the enterprise |
| Decide / Govern |
Clear definition of roles, responsibilities, and authority on the platform |
| Act | An AI Agent Fabric/Mesh architecture capable of integrating multiple agents and external AI |
Figure 5: Management Execution Platform Architecture for the Agentic AI Era
The diagram illustrates how the three shifts are implemented on ServiceNow.
“From passive tools to active collaborators,” “fundamental process transformation,” and “CEO-led enterprise strategy” are all realized on a single platform.
6. [Reference] Requirements for a Federated AI Agent Platform
Figure 6: Structural Requirements for an Enterprise AI Agent Platform
The diagram outlines the platform requirements needed for enterprise scale, their management implications, and how they are realized on ServiceNow.
Agent and workflow discovery, AI asset registries, observability, authentication and authorization, evaluation, feedback management, compliance, and risk management—these seven interlinked capabilities are prerequisites for operating Agentic AI in a manner that management can govern and control.
Conclusion: Competitive Advantage Belongs to Those Who Redesign Management for AI
The key question for CEOs is this:
Has our organization designed its decision-making, execution, and learning cycle with AI as a foundational assumption—or are we merely adding AI onto our existing management structure?
In the Agentic AI era, competitive advantage does not belong to the company that adopts the most AI tools, but to the company that redesigns its operating model with AI as a given.
ServiceNow provides the foundation to materialize that redesign as something management can deliberately design and control. What is required now is not the expansion of PoCs, but a transition to an AI-native operating model.
*This article draws primarily on McKinsey QuantumBlack, “The Rise of Agentic AI: A New Strategy for Enterprise Transformation” (August 2025). The views expressed are those of the author and do not represent the official position of ServiceNow.
