Thoroughly digest this AI value formula, and turn AI from a cost item into a growth leverage.

Many enterprises' AI investments fail to deliver expected returns. The problem does not lie in technical failures or inadequate change management, but in the most value-leveraging variable — the fit between AI and corporate strategy, which is often overlooked. Strategy, in essence, is about choosing which problems to solve, and the same applies to the AI field. Invest capital in areas that can generate tangible value.

Nowadays, enterprises across all industries are having heated discussions about AI. Executives are rolling out pilot projects, organizing employee training, upgrading data platforms, and chasing the latest AI models. But when they talk to CFOs about the financial results brought by AI, the conversation often falls into awkwardness. Enterprises are actively investing in AI, yet it is difficult to quantify the corresponding returns.

The gap between AI-related initiatives and measurable financial returns has become the core challenge faced by AI project leaders in large enterprises. A global survey released by PwC earlier this year shows that most CEOs stated that the application of AI in their enterprises has not brought quantifiable benefits.

We have observed this phenomenon across a wide range of industries: One of the authors of this article, David, has provided consulting services for enterprises in the financial services, consumer goods, and industrial sectors; the other author, Krishnan, has worked for many years in the capital-intensive oil, gas and chemical industry, where a misstep in a major strategic bet can lead to losses of billions of dollars.

Combining the practical experience of the two, we have identified the root cause of the AI value gap, and sorted out how managers can break the deadlock to maximize the benefits brought by AI.

The AI Value Gap

When AI fails to deliver measurable value, most managers assume the problem lies in technology or personnel. But based on our experience, underperformance against expected returns is essentially a strategic issue.

To put it bluntly, enterprises often apply AI to low-value opportunities. We have witnessed a large number of such cases firsthand:

- Solve internal frictions that are visible to the management but do not actually cost the enterprise much;

- Add new features that are irrelevant to customers;

- Copy mature application scenarios just because other companies have already done so;

- Present data in new formats without changing decision-making methods;

- Provide employees with tools that help them save time on tasks, but cannot bring quantifiable cost savings to the company.

Each of the above scenarios can only bring at best meager direct benefits to the enterprise, let alone build long-term competitive advantages.

When AI truly generates measurable value, it must be when the enterprise focuses on core issues related to profitability:

- Nike built an AI system to predict demand fluctuations, reduce inventory backlogs, and avoid stockouts — two factors that directly affect cash flow and customer satisfaction;

- Netflix uses AI to optimize content recommendations, which impacts user viewing duration and churn rate;

- Nestlé leverages AI to accelerate end-to-end product innovation, bringing new products to market ahead of competitors;

- Merck applies AI to reduce "misjudgment scrapping" in the production process — the situation where products are incorrectly eliminated during production, delaying the delivery of life-saving drugs.

The AI Value Formula

In enterprise scenarios, we can define AI value with a simple formula:

Value = Technology × Adoption × Strategic Fit

The first element "Technology" includes AI models, infrastructure, computing power and data. The second element "Adoption" is the core goal of change management, which David has specifically discussed in his recently published book. The third element "Strategic Fit" measures how well the problems you use AI to solve align with the company's strategic priorities, balance sheet and income statement.

The key to this formula lies in the multiplication sign: The three elements are multiplied, and as long as one of them is close to zero, the overall value will be zero. Just imagine three scenarios: the model is excellent, but there is a lack of suitable data; the solution is exquisitely designed, but no one is willing to use it; the tool is powerful, but it solves irrelevant problems. In any of the above scenarios, AI cannot create value for the enterprise.

The Overlooked Element: Strategy

Theoretically, the three elements are equally important. But in practice, we find that strategic fit is the most easily overlooked, and at the same time the link that leverages the greatest value potential.

Most enterprises have realized the importance of the first two elements: train appropriate models with correct data, and promote employees to embrace new working methods; but they ignore strategic fit. There are many reasons for this: AI projects are assigned to IT or innovation teams that are not responsible for profit and loss; managers assume that as long as a powerful model is connected to all data, there is no need to make difficult choices about which business problems to solve; limit AI goals to training and implementation, as if it is just a skill to be learned; or the entire project starts from the technology itself, purely out of fear of being left behind by competitors.

When we fix the two variables of technology and adoption, and only change the business problems that AI aims to solve, we can intuitively see the huge role of strategic fit.

For example, there is a solution that can help employees save one minute on routine work. In one enterprise, this one minute is worthless: the saved time becomes idle working hours, this task is not in the key process, and no changes will occur in subsequent links. In another enterprise, the same saved one minute can bring disruptive changes. The reason is that this task is in the bottleneck link that restricts production capacity (such as additional loading of goods to be transported to trucks), or employees need to repeat this operation tens of thousands of times a day, or this one minute reduces the time employees are exposed to hazardous environments. The same "optimization improvement" can lead to vastly different final values. The difference comes entirely from the strategic level: it depends on the position of this task in the enterprise's profit logic and risk chain.

Invest Capital in Areas That Can Generate Value

To leverage value through strategic fit, every AI investment must be anchored to the fundamentals of enterprise operation — invest capital in areas that can truly generate value.

Specifically, it can be divided into four steps:

1. Define Value Drivers

Value drivers refer to levers that, once adjusted, can tangibly change the profitability of an enterprise.

Take the heavy industry as an example. Procurement is a core value driver, and procurement expenditure can account for 30%–60% of the enterprise's annual revenue. The enterprise's procurement process may connect with hundreds of suppliers, and each supplier's quotation includes hundreds of detailed cost items, accompanied by terms of service agreements. It is impossible to optimize each item manually. However, some enterprises have successfully built an AI platform to optimize the negotiation process, reducing procurement costs by 10%–20%, which directly affects earnings per share.

Asset utilization is another major value driver in heavy industry. Reducing unit production costs is critical, and ensuring continuous operation of factories is a top priority. In the Permian Basin of the United States, leading oil and gas enterprises have launched AI drone projects: based on preset operation standards, it determines whether on-site personnel inspections are required. This predictive maintenance solution cuts unplanned downtime by 30%–40%, saving hundreds of millions of dollars in costs.

Other industries have their own core value drivers: for B2B service enterprises, optimizing sales team returns is the key; for subscription-based media enterprises, the priorities are customer retention and reducing customer acquisition costs; the service industry values labor productivity; retail enterprises focus on pricing optimization and inventory management; in the packaged consumer goods industry, drivers include product creativity, product portfolio management, and marketing mix optimization.

2. Identify Your Position in the Industry Value Chain

Value drivers vary greatly across industries, and even within the same industry, different enterprises have different drivers, which largely depends on the enterprise's position in the industry value chain.

Take chemical manufacturing as an example. Celanese, a bulk commodity producer, mainly produces chlorine and polyethylene; while Givaudan is a specialty chemical producer that produces flavors, fragrances and cosmetic raw materials. Both belong to chemical manufacturing, but their value drivers are completely different.

Celanese has invested billions of dollars in building large-scale production facilities such as steam crackers to produce chemicals traded in the global market, and its value is driven by compressing costs through operational efficiency.

Givaudan produces high-margin products, and its R&D investment as a percentage of revenue is 5 to 10 times that of Celanese. Its profits come from intellectual property, with a shorter innovation cycle. Its value driving priorities lie in the speed and cost of new product development, as well as the ability to deeply embed itself into customers' production processes.

3. Look for Opportunities in the Income Statement

Next, turn your attention to the income statement. Go through the income statement of the company or business unit line by line, and think: Which item can AI change? How much difference can it make? For revenue, cost of goods sold, and operating expenses, imagine the possibility of creating value for each account.

Take a large oil and gas company with assets concentrated in upstream businesses as an example. Analyzing the income statement reveals that reserve life and low-cost production are critical. The corresponding AI directions are first to reduce production downtime to ensure revenue, and second to optimize exploitation to extend the utilization cycle of reserves.

The income statement can also help enterprises tap into new revenue sources and revitalize the capital they have already invested. For this oil and gas enterprise, it can use AI to carry out capital leasing, or use spot bidding, futures and options to conduct crude oil spot trading to open up new revenue streams.

4. Set Financial Return Measurement Indicators

Finally, you need to determine how to measure the impact of AI on the selected value driver, that is, the target problem to be solved.

Merck measures the misjudgment scrap rate in the production process, as well as the product capacity available to patients; Netflix's indicators include user retention, viewing duration, and long-term satisfaction, paired with rigorous A/B tests to evaluate the effect of the AI recommendation system; Nike tracks the inventory mismatch rate (both backlogs and stockouts), the holding cost of excess inventory, and the accuracy of demand forecasting in the new product launch phase; Nestlé's indicators focus on the speedup of each R&D link and the new product launch cycle.

Indicators must be determined before building the AI solution and integrating it into the business process. If you cannot define the measurement method of the project in advance and who will confirm the value effect, then you are not ready to launch the project.

Impacts of Anchoring AI to Corporate Strategy

Making all AI projects closely aligned with corporate strategy is the key to turning investments into quantifiable returns. We have observed that enterprises that adhere to this set of strategic principles have avoided the trap of "only doing things but no returns", and can implement AI on a large scale to achieve measurable results.

Saudi Aramco, where one of the authors Krishnan once worked, disclosed that AI technology has brought rich returns, with earnings reaching 2.6 billion US dollars in 2025 alone. Relying on its AI recommendation system to influence users' daily viewing behavior, Netflix reduces annual losses related to customer churn by 1 billion US dollars. Leveraging its self-developed AI platform HawkAVI, Merck has reduced the misjudgment scrap rate of products across multiple product lines by more than 50%. After Nike launched its AI forecasting model, the inventory mismatch rate decreased by 20%, the stockout rate dropped by 15%, and it increased sales by 10% in target regions through demand forecasting. Nestlé uses AI to shorten the product creativity cycle, reducing the pre-development ideation time of new products from six months to six weeks.

Anchoring AI to value brings far more changes than improving the performance of a single project, it will also reshape the positioning of AI within the organization. When each project is tied to value drivers and income statement accounts in advance, managers responsible for profit and loss and investment decisions will quickly gain confidence, which in turn drives the whole company to build confidence.

At the same time, the competitive advantage built by the enterprise will also transform: from short-term scattered results to sustainable advantages. General AI implementation cases provided by vendors cannot build a moat, as competitors can obtain exactly the same models and application scenarios. But when AI is deeply integrated with the enterprise's business operation logic, the situation is different: strategy itself is a moat, and AI becomes a tool to implement this strategy.

The value formula we proposed is a simple and practical tool to help AI projects find the right direction and realize commercial monetization. Many enterprises' AI investments fail to deliver expected returns. The problem does not lie in technical failures or inadequate change management, but in the most value-leveraging variable — the fit between AI and corporate strategy, which is often overlooked. Strategy, in essence, is about choosing which problems to solve, and the same applies to the AI field. Invest capital in areas that can generate tangible value.

David L. Rogers, Krishnan Sankaranarayanan | Article

David L. Rogers is a professor at Columbia Business School. Krishnan Sankaranarayanan is the Head of AI Transformation at Eastman, with 24 years of industry experience. He has previously worked at ExxonMobil, SABIC and Eastman, and has rich practical results of value implementation.

This article is from the WeChat Official Account "Harvard Business Review" (ID: hbrchinese), Author: HBR-China, republished by 36Kr with authorization.