——In the AI Era, Organizational Restructuring Is More Urgent Than a Technology Race
BeijingJune 15, 2026 /PRNewswire/ — Recently, at the 28th Beijing International High-Tech Expo, Dong Haijun, Partner and General Manager of Strategy and Transformation Consulting, IBM Consulting Greater China Group, delivered a speech sharing insights on the current phenomena and challenges of enterprise AI transformation, the underlying causes, and directions for action, engaging in discussions with attendees.
Dong Haijun, Partner and General Manager of Strategy and Transformation Consulting, IBM Consulting Greater China Group
Foreword:
Phenomenon and Root Cause: We often see a contradiction—”AI can launch an intelligent process in three days, yet struggles to add a single cent of profit to the financial report.” The root cause behind this is not a failure of technical capability, but a lack of organizational change capability.
Direction for Action: AI transformation, in essence, is a concentrated effort to make up for historical debts in processes, data, and organizational capabilities. It is not a reset of the “starting line,” but a systematic settlement of existing “debts.”
As Arvind Krishna, Chairman and CEO of IBM, pointed out: “True leading companies are not deploying more AI tools, but redesigning their own business operating models.” What enterprise-level AI applications require is precisely a new set of operating models—this is the key to whether transformation can be realized.
The following content is compiled from the speech and used with relevant authorization.
Over the past three years, every Spring Festival has felt like a technological tipping point. 2024 was Sora, 2025 was DeepSeek, and this year it’s “Lobster.” In some companies, Agent applications are being deployed at a rate of hundreds per day. IT departments are busy reporting deployment numbers, and employees are eager to try new tools. However, in conversations with dozens of CEOs, we hear a common confusion: Money has been spent, people have been mobilized, so why hasn’t profit increased? Why hasn’t competitiveness changed? AI is booming on the front lines, but top management sees no measurable real value.—This is the paradox of current AI transformation.
70 Years of History Clarify One Thing: AI Is Still in the “Warm-Up Phase”
Looking back at history: Every major technology—whether the internet, computers, mobile phones, or data analysis—takes an average of 50-70 years from inception to commercialization, going through three main stages: early technical feasibility research → commercial feasibility → ecosystem maturity. AI research began in the 1950s and has only recently started to truly enter the commercial phase. The widespread discussion today about “scaling AI applications” precisely indicates AI’s current stage: transitioning from “technically feasible” to “commercially feasible,” a relatively early phase.
Without this historical perspective, one misjudges the essence—today, AI is at the starting area of a cross-stage transition, still in a “warm-up” nature. However, AI inherits the consistent logical thread of the information revolution: the continuous liberation of human labor.
We have already completed three clear advancements: liberating computing power, liberating recording capabilities, and liberating analytical support for decision-making. Next, AI liberates “cognitive and autonomous decision-making” capabilities. It is in this fourth advancement that enterprises encounter many unprecedented difficulties.
Root of the Dilemma—The Gap Between Technical Capability and Organizational Capability
“Trying to force artificial intelligence into an existing organization is very likely a wrong approach.”—The sentiment expressed by the CEO of Philip Morris International echoes the experience of many corporate executives facing the current common dilemma of “AI booming, returns flat.”
The first three phases of informatization liberated deterministic capabilities—calculating faster, recording more accurately, analyzing more comprehensively. The fourth advancement essentially involves AI liberating non-deterministic capabilities—participating in judgment and decision-making. This change creates a sharp conflict with existing organizational structures designed for deterministic, linear processes.
What seems like a problem solvable through technology stacking actually creates two typical mismatches. First, value mismatch: AI is used for low-value tasks like writing emails, summarizing, and organizing documents, without intervening in core decision-making, customer relationships, or supply chains. The savings are invisible on financial statements. Second, power mismatch: AI systems developed by IT departments are rejected by business units. Why? Because they “steal jobs.” A business professional with 20 years of experience sees IT create an AI in three days claiming superiority—who would accept that?
In the real world, the real challenge is not technology itself, but how to enable collaboration between humans and technology, preserving accumulated experience while gaining new efficiency and business value through technology.
Solution Approach—Fragmentation Is the Disease, Systematization Is the Cure
Data shows that 74% of enterprises have implemented point-based AI applications, yet struggle to link them to real business value.
Now, with AI, changing a fourth- or fifth-level process takes only three days—this is “mobile warfare,” like guerrilla operations: rapid iteration, small steps, fast progress. It’s easy to start, but problems follow: point-based successes cannot automatically aggregate into systemic change.
After shifting from “positional warfare” to “mobile warfare,” a stark contradiction emerges: the easier it is to start, the harder it is to scale. Fragmented Agents have disconnected data, lack dialogue with each other, and misaligned interests.
This explains why today one Agent can generate another, and succeed 90% of the time. But “point-based sparks” don’t translate into business value or measurable ROI. Abandoning the obsession with “individual AI efficiency” and shifting toward systemic capability reconstruction is the choice enterprises need to make today.
Scaling doesn’t lack technology; it lacks systemic capability reconstruction. This brings us back to the fundamental question: With a three-year horizon, how should enterprises think about and drive AI transformation?
The Secret of Change—Prosperity Doesn’t Foster “Change-Ready” Organizations
On this question, during a visit to the Autodesk CEO, I received an unforgettable answer: “You can’t predict every peak, but you can prepare by building an organization that is resilient, adaptable, and ready for change.”
I once deeply served a leading company, supporting its successful transformation. IBM’s value was not just technical; the decisive factor was organizational capability building, enabling the company to self-evolve. An organization’s core capability is not efficiently executing a set strategy, but continuously self-reshaping and leaping in uncertain environments.
IBM itself exemplifies this logic. Founder Thomas J. Watson Sr. and his successor Thomas J. Watson Jr. pioneered an 85-year “heroic era,” where their genius vision and extraordinary will drove rapid growth. But in the 1990s, IBM fell into crisis, forced to re-examine its organization, processes, operations, and future value. Change agent Lou Gerstner took over in a critical moment; his most valuable legacy was not saving the company, but instilling a gene for change. Since then, IBM has undergone three major transformations: process transformation, data transformation, and the organization-wide AI-driven transformation launched in 2020, ongoing for five years. IBM is its own “AI Zero Client.”
Changing oneself is always harder than sprinting on the same track! But capabilities don’t come for free. IBM’s AI transformation delivered results: the original target of $2.5 billion in financial benefits was ultimately achieved at $4.5 billion. To date, IBM is the only traditional IT company globally to successfully transform into an AI company, ranking among the top five global AI companies with a century-long history.
The Truth of Transformation:AI Transformation Is Not a Starting Line Reset, But a Settlement of Three Types of “Debts”
In January this year, at a global partner AI transformation discussion in New York, we found that foreign companies started AI transformation earlier, not because they were smarter, but because their foundational groundwork was better.
As early as the 1990s, foreign companies had begun large-scale process transformation. In 1994, two-thirds of large U.S. companies had initiated process reengineering. Domestically, it was significantly later. Taking one of our well-known electrical appliance company clients as an example, it only started process reengineering in 2007 and completed the “one company, one system, one standard” transformation in 2012.
Why are processes so important?—Without processes, there is no good data; without good data, AI cannot automate smoothly.
We now realize that the endgame was AI all along. Over the past 70 years, putting data on the internet and embedding business procedures into IT systems were all preparations for today’s AI takeover. For AI to take over, enterprises must have a solid foundation in data, processes, and organization. Historical deficiencies in these areas are called “AI debts.”
In the first three phases of informatization, each step left behind corresponding “debts.” Even in the last three-plus years, we have still been helping many enterprises with process changes. IBM’s own experience reveals a pattern: the smoother the transformation in the previous phase, the faster the transformation in the next. In AI transformation, a concentrated settlement of three types of “debts” is necessary.
Solution Method: Three Levers, One Sequence
Based on IBM’s own transformation experience and service practices with hundreds of enterprises, we have gradually derived three core levers for driving AI transformation: Leadership → Work Model → Technology Implementation. The sequence of these three levers cannot be reversed.
Lever One:CEO Leads, Business Transformation First
Any AI transformation project must be led by the highest level. This is not just a slogan, especially in settling “organizational debt”—it is decisive.
Why? Because business units naturally feel it “steals jobs.” As early as 2017, during an internet transformation project for an airline company, we had a discussion predicting a problem: In a VIP lounge, if a newly hired young woman wearing AI glasses greets guests, would it cause discomfort for employees who have worked there for 20 or 30 years? The answer was yes—not from defiance, but from anxiety. Decades of experience could be leveled by a pair of glasses. This small example tells us that projects must be driven by top leadership, with business transformation leading the way—first solving how business reshapes its competitiveness in the AI era, then applying IT and AI expertise, not the reverse.
The CEO leads the formation of a “Transformation Steering Committee,” while also establishing a “Productivity Exploration Team.” This is a highly comprehensive team, including business experts, data engineers, algorithm experts, legal advisors, security officers, compliance managers, AI ethics committee members, and others, all of whom must be senior experts in their fields. This team is the backbone for addressing current point-based AI application challenges and promoting scaled deployment—breaking down cross-departmental barriers, introducing external benchmarks, identifying specific improvement opportunities, while fully empowering every employee to achieve process simplification and automation. This system achieves seamless alignment from strategic decision-making to employee action, ensuring effective transformation progress.
Core Principle: Leadership drives change, settling “organizational debt.” Business drives technology, not technology driving business.
Lever Two: Reconstruct Work Models, Focus on Four Entry Points
Reconstructing work models is complex and systematic. It can start from four areas, delving into enterprise operations to settle “data and process debts” while advancing AI transformation.
First, internal and external data insights and decision support. Managers spend a lot of time daily reading reports, spreadsheets, and attending meetings. AI’s most direct value is automatically analyzing, organizing, extracting, summarizing, and guiding data. Freeing people from information overload. IBM prioritizes this as its first priority.
Second, personalized customer interaction reshaping. AI can achieve personalization at scale. IBM helped a Swiss bank compare human agents with AI agents, finding that AI agents were better at judging tone—when a customer was unhappy or impatient, they proactively transferred to a human agent. This type of human-machine collaboration is highly valuable.
Third, IT modernization itself. Essentially, AI is coding, and the easiest thing to code is coding itself. IBM applied AI to the software development lifecycle (SDLC), reducing the IT team size by approximately 80%, freeing up significant gains.
Fourth, employee productivity and experience enhancement. AI provides new interaction methods, freeing employees from cumbersome content management systems. Take IBM AskHR, which has been running for nearly five years. Employees can request leave, obtain employment certificates, or change work locations through natural language conversations. While seemingly minor, these applications are the most direct window for employees to perceive AI’s value daily.
Core Principle: Not a blanket rollout, but focus on high-frequency, quantifiable value anchors. Only when data flows and processes are smooth can AI truly run.
Lever Three: Technology Implementation, Starting from 0.1
AI is different from all previous technology applications. In 2018, IBM’s Chairman proposed building an “agile culture”: without an agile culture, AI applications are impossible. Why? The logic of action for an AI-oriented organization is completely different:
- Traditional software (e.g., ERP): Launch must be version 1.0; launching version 0.1 is a disaster.
- AI applications: Always start from 0.1. Without 0.1, there will never be 1.0; it grows through use.
Accepting errors, failures, and tolerance is not just a cultural issue but a system-building issue. A fault-tolerant mechanism is a prerequisite, and enterprises have “organizational debts” to settle here. Of course, optimizing incentive mechanisms, valuing technical skills, and strengthening recruitment and retention are important alongside technology.
Core Principle: Agile culture + Fault-tolerant mechanism = Soil for AI implementation
Turning AI into real profits for the enterprise is not simple. It could even be said to be more difficult than all transformations of the past 40 years combined. Where is the difficulty? It boils down to implementing three key actions.
Implementation Points and Safeguards: Levers into Action, AI into Real Value
First, set a “North Star metric” with clear financial targets. Without it, enterprises cannot sustain transformation amidst numerous difficulties, resistance, and criticism. There must be a quantifiable, guiding clear metric; it is the only compass for mobilizing the organization.
Second, establish an AI governance mechanism. Clarify the AI accountability chain. Previously, when implementing IT systems, responsibility was clear. Now, if AI goes wrong, how is accountability assigned? A wrong AI data point can cause catastrophic consequences. Whose responsibility is it? The business unit? The data scientist? The algorithm engineer? Or the user? We now have many methods, such as establishing an Agent committee, multi-system collaborative review, etc. This path is full of trial and error but cannot be bypassed.
Third, repeatedly remind of the historical truth—In the transformations of many enterprises over decades, every time a new technology arrives, everyone is excited, believing it’s a new starting line to overtake competitors on curves and re-energize with new tech. But history repeatedly proves: New technology does not reset the starting line; for leaders, it’s an accelerator; for laggards, it means more “debts” to repay and catch up on. Today’s AI is no exception.
AI transformation is not a technology race, but two sides of the same coin: the concentrated settlement of enterprise “debts” and the self-reconstruction of organizational capabilities.
Media Contact
IBM Corporation logo.