IBM Consulting's Hou Junwei: AI reduces the work, so why isn't the money rolling in?

BEIJING, Oct. 30, 2026 /PRNewswire/ -- As businesses become increasingly software-driven, their core capabilities increasingly run on application systems such as ERP, CRM, supply chain, and HR. Application systems have evolved from back-office support tools into core platforms that carry customer service, operations management, risk control, and business innovation, and their operational quality largely determines the operational quality of the enterprise.

Driven by AI, Application Management Service (hereinafter referred to as AMS) is undergoing a fundamental transformation: from focusing on the continuous management and optimization of "the application systems themselves" to predicting and preventing problems and automating workflows, and further to focusing on "the enterprise capabilities carried by the application systems." There is still a clear gap in this leap.

Most enterprises (91%) have grasped this trend and expect to adopt "AI-enabled AMS" within a few years, shifting from "keeping systems running normally" to "driving the continuous evolution of enterprise capabilities," but only 3% have truly made progress. Roles are upgrading, yet action is slow; the crux is not a lack of AI technology, but that the path from "AI efficiency gains" to "measurable value" has not yet been paved.

To this end, the IBM Institute for Business Value, in collaboration with Oxford Economics, conducted a survey of 1,000 enterprise executives worldwide and found that if three key points are handled properly and aligned with the way AI creates value, AMS can upgrade from "application management" to a "performance engine." These three points correspond to three changes.

Change the compensation formula --

Break out of "AI reduces work input x priced per person-day = the higher the efficiency, the lower the revenue"

For decades, the economic logic of AMS has been built on "more work creates more value." Currently, as many as 86% of contracts still measure services and calculate compensation based on labor input, ticket volume, and completed workload. AI precisely reduces labor input; compared with technical capability, more than 70% of enterprises (as many as 76% of leaders in China) consider workload-based compensation contracts to be a more prominent constraint, because they turn a decline in labor input into a decline in revenue, which is a significant obstacle to advancing AMS. The related incentives, governance, and operating methods still follow the old logic designed for manual execution, and the application management service evaluation system suffers from a structural imbalance.

The old model is unsustainable. The AMS operating model must be proactively restructured with AI in order to keep pace with the new reality, including: focusing on and measuring reliability, business resilience, and failure recovery capability; anchoring compensation to the degree of improvement in business outcomes; and tying incentive schemes to problem prevention and continuous business evolution. 62% of Chinese enterprise leaders expect that by 2028 enterprises will adopt such an operating model, leaving room to unlock high value at scale.

Change the way work is done --

Choose the best entry point + workflow design + embedded governance

AI-enabled AMS is the direction, but rolling it out across the board from the start is not wise -- it requires large investment, delivers slow results, and concentrates risk. The most pragmatic approach is to advance the transformation without sacrificing operational stability. Start with areas where "reliability pressure is greatest and outcomes can be quickly verified." Recurring failures, high-frequency service requests, and manual processing bottlenecks are the best breakthrough points: they can deliver performance improvements fastest and also build confidence for subsequent scaling.

However, the results of the "entry point" must be solidified, and turning a single point into scaled results depends on redesigning workflows. The entry point proves effectiveness, while workflow redesign makes effectiveness the norm. The value of AI is reflected through integration into daily work -- teams identify risks earlier, analyze root causes, and take action; operations shift from reactive problem-solving to proactive prevention, reducing the time consumed by recurring failures. In fact, nearly 80% of surveyed Chinese enterprise leaders say they are advancing adjustments for this purpose.

It should be emphasized that at the scaling stage, governance must be embedded into workflow design: clarify the boundaries of responsibility such as the resources AI agents can access, the operations they can perform, when to pause and escalate, and the steps requiring human oversight. These safeguard mechanisms provide reliability assurance for AI-enabled AMS operations. In addition, build in a continuous optimization mechanism to regularly review workflow design, model performance, and the governance framework to ensure continuous improvement.

Change where resources flow --

From investing in "systems" to investing in "the capabilities systems carry," from "efficiency gains" to "performance gains"

The earliest returns of AI-enabled AMS appear as enhanced reliability: faster detection, better recovery, and smaller business impact from incidents. 69% of surveyed managers say this is only the first layer of returns; the returns of the new model go beyond efficiency. Improved reliability frees up operations resources that can be used to support the continuous evolution of applications and, over time, accumulate and strengthen into a return dividend, with returns gradually extending to the business level.

The value of application management is not just that "systems run steadily," but also in creating conditions for "the business carried by the systems" to move forward. The freed-up resources and capabilities are used to accelerate the modernization process -- simplifying applications, improving performance, and advancing modernization. The way AMS teams work changes accordingly, shifting from maintaining performance to business evolution and value conversion.

But the report also warns: early returns can advance the transformation, but they usually cannot sustain themselves. If the freed-up capacity is again consumed by day-to-day support work, the results remain at "efficiency gains" rather than "performance gains." Leading enterprises do not channel it back but invest it in capabilities, continuously directing the freed-up resources toward application modernization, preventive capability building, improved observability, and the evolution of AI applications, thereby amplifying business value.

From ensuring applications run to ensuring that enterprise capabilities continue to evolve, "efficiency" is thereby converted into "performance."

Three actions to put this judgment into practice

The three changes answer what to change; the specific starting points can be briefly summarized as three actions:

1. Restructure contracts so that service returns are tied to outcomes

Re-examine the assumptions in contracts based on manual labor volume, and link incentives to reliability, availability, recovery performance, incident reduction, and business impact. At the same time, establish a concise set of outcome metrics and use the same set in contracts, governance, and performance reviews, rather than each party speaking a different language.

2. Before expanding AI applications, first change the way work is done

First identify bottlenecks such as high-frequency failures, repetitive requests, and manual handoffs, so that initial AI applications directly address the root causes of the problems rather than bypassing the pain points to build generic tools. Operations capacity should gradually be allocated to proactive prevention, reliability optimization, and workflow restructuring, with improvement results measured against clear baselines.

3. Set a clear direction for reinvesting the freed-up resources and capabilities

Establish a dedicated list of improvement and modernization tasks, track where the freed-up capacity is actually used, and set reinvestment targets linked to reliability, modernization, and operational performance. Where the saved effort goes must be accountable.

The principles are common, but enterprises differ and so do their paths

The "three changes" discussed above simply answer three questions under the AI-enabled AMS model: How is money calculated? How is work done? How are the saved resources used? The "three actions" above are implemented by following the way AI creates value, making application management a performance engine.

How can this judgment be turned into reality? IBM's consulting practice is to establish a new type of partnership with clients and advance collaboratively across multiple dimensions:

- In next-generation talent development, enhance the team's IT skills to support innovation;

- In organizational integration, advance the review and optimization of organizational structure based on business objectives, integrate processes, iterate mindsets, and promote workflow standardization;

- In cost optimization, take stock of and make IT spending transparent, integrate resources, implement automation, and advance fixed-cost optimization.

A large communications service provider is one such example. Its entry point was not organization or cost, but the platform's intelligent capabilities themselves. The client hoped to rely on large-model capabilities to connect internal service scenarios across the entire IT lifecycle, build an integrated AMS platform covering intelligent customer service, automated inspection, and automated fault handling, and improve operations quality, autonomy, and scalability.

IBM helped the client restructure the platform's intelligent capabilities in three areas:

First, build an intelligent Q&A bot to handle routine inquiries, consolidate service knowledge from dozens of business lines such as government and enterprise and sales, as well as high-frequency questions from phone and WeChat channels, into a unified knowledge base, and then use policies to guide users to prioritize self-service queries, improving the productivity of frontline customer service;

Second, replace manual inspection with automated inspection, automatically monitor server status based on threshold rules and large-model comprehension capabilities, and after identifying faults such as CPU anomalies, automatically trigger handling processes such as tickets and emails. At the same time, rely on the large model's flexible understanding to reduce the development of prebuilt rules and lower iteration and deployment costs;

Third, connect full-link data, establish ticket funnel analysis, track the proportion of AI-initiated versus manually initiated tickets, and continuously optimize automation strategies, with the goal of raising the AI ticket share to 60%–70%.

After the project was implemented, overall IT operations costs fell by 50%; the intelligent Q&A bot can directly handle at least 40% of daily inquiry volume. The platform also formed a reusable large-model asset foundation, accelerated enablement of platforms such as low-code, and expanded to multiple business lines such as finance, human resources, and supply chain, supporting intelligent customer service in multiple domains. Relying on natural language interaction and large-model capabilities, the platform is moving from stacking conventional functions to data-driven intelligent services, maximizing the value of asset reuse.

Conclusion:

From ensuring applications run to ensuring that enterprise capabilities continue to evolve—the efficiency brought by AI can truly be converted into performance only through the three changes of compensation, ways of working, and resource flows. Methods can vary by enterprise, and paths can have different emphases. Only when money is calculated correctly will teams be willing to work differently; only when work runs smoothly can saved resources be invested; only when resources are invested correctly will efficiency accumulate into sustained business value. In the final analysis, this is the new economic logic of AI-enabled AMS—efficiency is not the endpoint, but the starting point.

To learn more about China insights and action recommendations, please read the IBM Institute for Business Value report "Application Management: The Enterprise Performance Engine."

Data note: This report is based on a survey conducted by the IBM Institute for Business Value in collaboration with Oxford Economics, covering 1,000 senior leaders across 13 global markets and 21 industries. China data in this article all refer to Chinese respondents or respondent organizations in the report; 2028-related data are respondent organizations' expectations.