According to an August 14 report by Yicai, Wang Puzhong, CEO of Meituan’s Core Local Commerce division, discussed his views on enterprise AI adoption in a public speech and shared details of Meituan’s internal AI transformation.
Wang said that although AI offerings are abundant, only a small number of companies can currently quantify measurable returns from real-world deployment. He attributed this to four mismatches: in perception, companies either exaggerate or underestimate AI; in efficiency, large models are used for routine tasks without reaching core operations; in use cases, AI remains on the periphery rather than being integrated into the heart of the business; and in assessment, companies seek immediate results while overlooking their degree of internal readiness. In Wang’s view, these four overlapping mismatches make it difficult for AI investment to translate into measurable productivity gains.
Wang revealed that Meituan’s AI transformation has gone through four stages. The first stage ran from February to March this year, when the company launched a companywide “shrimp-raising” campaign. However, bills soared, with daily spending exceeding RMB 10 million, while hallucinations produced during the campaign disrupted actual business operations.
In the second stage, beginning in April, each Meituan business unit established an AI organization. The company subsequently developed an effective set of transformation steps to systematically convert employees’ enthusiasm into productivity.
By June and July, Meituan’s AI transformation had entered its third stage. Through a competitive “horse-racing” mechanism, the company refined its methodology and reached a key conclusion: AI transformation is not merely a technical issue, but a systematic undertaking integrating business, organization, and technology, and cannot rely solely on product and R&D teams. By July, AI had initially been successfully integrated into Meituan’s internal product workflows and begun generating value.
Meituan’s all-scenario AI Agent platform, CatPaw, officially launched in July this year and has already been deployed at scale within the company: it has reached 90,000 employees, supported the creation of 30,000 Agents, and been validated across multiple real-world business scenarios.
CatPaw offers a standalone mobile app and PC client, with real-time task synchronization and seamless collaboration across both platforms. The mobile app allows users to initiate tasks at any time, check progress, and remotely confirm critical decisions, while the PC client focuses on in-depth local execution and offers comprehensive capabilities including file operations, browser control, and terminal commands.
CatPaw’s cloud mode supports uninterrupted 24/7 operation, even when local devices are powered off or disconnected from the internet. Long-running and scheduled tasks continue operating in the cloud, and users can view the results at any time afterward by opening the app on their phone or computer.
Building on its general-purpose, all-scenario AI Agent capabilities, CatPaw also incorporates Meituan’s end-to-end industry knowledge in local services. Its extensive expertise spanning store operations, review systems, marketing conversion, order fulfillment, and delivery has been packaged into ready-to-use experts and skills. These cover use cases including store review diagnosis and optimization, product copy generation, marketing material design and evaluation, campaign planning, and business data analysis.