Lao Gan Ma Uses AI: Costs Saved, but Can It Keep the Flavor?

A food company that won the market with traditional flavors has begun using AI for quality control of raw materials—what changes will this bring? Recently, the news that Lao Gan Ma's 2025 revenue reached 5.4 billion yuan drew attention alongside reports that it had introduced an AI visual inspection system, and some news headlines even went so far as to claim sensationally that "Lao Gan Ma earns 540 million a year with AI."

Yang Wei, Associate Professor of Strategy and Entrepreneurship at CEIBS, offers a quick take: as more and more traditional companies begin to adopt AI, what is worth discussing may not just be how much cost can be saved or how many workers can be replaced, but more importantly, which tasks are suitable to hand over to AI, and which experiences and values still need to be upheld by people. She further cautions that a company using AI to improve efficiency does not mean consumers will perceive better product value.

Lao Gan Ma's operating revenue already reached 5.26 billion yuan in 2022, 5.391 billion yuan in 2024, and about 5.4 billion yuan in 2025—little changed. The AI visual inspection system was fully rolled out in 2025, but no public data yet shows how much revenue it contributed or how much of a performance leap it achieved.

AI's more direct effect may show up in quality management. According to Li Xin, Executive Vice President and Chief Quality Officer of Lao Gan Ma, after the AI visual inspection system was fully launched in 2025, the defect rate for key-process products dropped to 0.02%, and core raw materials became 100% traceable. However, public reports have not yet provided comparable data from before the upgrade.

How much of these changes actually came from AI, and how much from equipment upgrades and management adjustments? Is it worthwhile once investment costs are considered? The available information is not yet enough to answer. Still, the Lao Gan Ma case offers a window into how traditional companies are applying AI.

For a food company like Lao Gan Ma, whether it can preserve the taste consumers know depends to a large extent on raw materials. The quality of chili peppers is the first gate for product quality. This raw material, which requires large-scale procurement and processing, is directly tied to product flavor.

Lao Gan Ma once sparked a taste controversy by adjusting where it sourced its chili peppers. Media reports said the company at one point increased its use of peppers from Henan and other places, prompting consumers to discuss a "change in taste." Whether the change in taste was caused solely by the shift in origin remains debatable, but raw material management affects the product flavor consumers know and may also bear on the customer trust the brand has built over time.

Industrial vision and optical sorting are already used in food processing to identify defects and remove foreign objects based on features such as color and shape. Companies still need on-site testing to determine whether the equipment can adapt to different batches of raw materials and actual production speeds.

Even a small improvement in sorting accuracy, scaled up to production volumes, can generate considerable returns over time.

Although it cannot simply be concluded that consumers prefer the latter, the associations evoked by the production method do affect consumers' impressions of a product.

A study published in the Journal of Marketing in 2015 found that, in its research context, labeling products as "handmade" increased their appeal, partly because consumers felt the maker had put care into them. Although this study cannot directly prove that chili sauce consumers have the same preference, it still offers some reference value.

Some people associate handmade production with careful ingredient selection and home-style taste, while others may feel machine sorting is cleaner and more consistent. Taken to an extreme, AI may instead trigger associations with "technology and harsh additives." So whether AI adds or subtracts points depends on how consumers understand it and whether the actual product has changed.

For a brand that has been through raw material controversies, consumers may be especially sensitive to cost cutting. If a company only emphasizes how much labor it has saved without explaining how quality is guaranteed, it is easy for people to worry that the familiar taste will change.

Human involvement may also have brand value in certain products and scenarios, but that does not mean human labor is necessarily better than machines.

What companies see as a cost issue is sometimes the very quality issue consumers care about most.

Lao Gan Ma previously adjusted its chili procurement sources to find room to cut costs, and now it has introduced AI into the raw material sorting stage—both times the focus has been on chili peppers. From a business logic perspective this is not surprising: chili peppers are Lao Gan Ma's core resource, and its ability to select and process peppers is an important foundation of its core competitiveness.

The pain points of business costs and the sensitivities of consumers thus converge on the same link. What companies see is room to improve efficiency; what consumers care about more is whether the taste will change. Both concerns are entirely reasonable.

But the trouble is that the two are not easily measured on the same financial statement. Changes in procurement prices and labor costs can be calculated quickly, while consumers' disappointment with taste may build up gradually and is hard to attribute to any single production adjustment. If a company focuses only on immediate costs, it may underestimate long-term brand damage.

This is also why past raw material controversies make it harder for a company to explain new changes.

What is truly worth preserving is not any fixed mode of production, but the flavor consumers know and their trust in the brand. Technology can be updated, but companies need to explain clearly what has changed and which quality requirements have not.

Around Lao Gan Ma's AI application, there is also a widely circulated figure: pepper-sorting workers were reduced from more than 600 to 48. But this set of numbers has not been verified. On October 26, 2026, multiple media outlets reported that Lao Gan Ma staff explicitly denied the related layoff rumors, saying the company had never released these figures.

Although the layoff rumors are unreliable, the question is still worth discussing: as machines take on more appearance-screening work, how should the experience accumulated by workers who have dealt with raw materials for years continue to play a role?

Skilled workers may rely on dryness, smell, appearance, and years of experience to notice that a batch of peppers is different from usual. Some of these judgments can be translated into inspection metrics, while others need further recording and verification. Whether machines can cover these judgments depends on the system's actual capabilities, and conclusions should not be drawn merely from the word "AI."

Rather than seeing AI only as a tool to replace human labor, Lao Gan Ma could also explore its role in R&D. Food companies looking for talent that understands R&D and is familiar with consumer needs are sometimes constrained by location and talent supply, and AI may just fill that gap.

Some food companies have already tried using AI to assist R&D. In 2019, seasoning company McCormick partnered with IBM to use data such as consumer preferences and sensory research to help find new flavor combinations, and launched the ONE series. According to reports, its R&D speed improved significantly.

Professor Profile

Professor Yang Wei is Associate Professor of Strategy and Entrepreneurship at CEIBS. Before joining CEIBS, she was an Assistant Professor of Management at George Mason University in the United States. Professor Yang graduated from the University of Texas at Austin with a PhD and a master's degree in management, and she earned a master's degree in finance from the University of Hong Kong and a master's degree in management from Peking University.

This article comes from the WeChat public account "CEIBS" (ID: CEIBS6688), author: Yang Wei, published by 36Kr with authorization.