Timothy Zhao

Timothy Zhao inquired about the detailed specifications and use cases of Tuya's upcoming AI home assistant, its anticipated impact on the overall business, and the effect of AI on PaaS and Smart Solutions segment growth, including volume, pricing, and gross margin.

Answer

Alex Yang, Co-founder and CFO, explained that the new AI assistant is designed as a comprehensive home assistant with multiple agents, capable of interacting with hundreds of millions of Tuya-powered physical devices to create a 'Jarvis-like' experience. This aims to lower the entry barrier for new users by simplifying interactions. Regarding AI's business impact, Yang noted that 93% of Q3 shipments included AI capabilities, indicating strong adoption. He stated that AI is seamlessly integrated across PaaS, Solutions, and SaaS, driving new demand and market penetration. While new AI features are expected to improve gross profit margins, the current focus is on market promotion and incubation.

Timothy Zhao from Goldman Sachs asked for more details on the AI home agent to be released at CES, including its specifications, use cases, and anticipated impact on Tuya's overall business. He also questioned AI's influence on the PaaS and Smart Solution segments, specifically regarding volume, pricing, and gross margin.

Answer

Alex Yang, Director, Co-founder, COO, and CFO of Tuya, clarified the new product as an 'AI assistant' designed to manage multiple home tasks, differentiating itself by interacting with physical devices across the Tuya ecosystem to simplify user experience. He noted that over 93% of Q3 shipments incorporated AI capabilities, indicating strong developer adoption. AI is seamlessly integrated into PaaS, solutions, and SaaS, expected to boost market penetration and eventually improve gross margins, though initial focus is on market incubation.

Timothy Zhao of Goldman Sachs asked about Tuya's competitive advantage in the IoT PaaS segment as the industry transitions to AIoT and how the company plans to maintain its edge. He also requested more clarity on the company's shareholder return policy, particularly regarding future dividend declarations.

Answer

Co-founder, COO, and CFO Alex Yang explained that Tuya is actively encouraging developers to adopt AI, highlighting that 93% of products shipped in the first half of 2025 already included AI capabilities. The company is also focused on incubating and commercializing innovative AI prototypes from its developer community. On shareholder returns, Mr. Yang confirmed that dividends are considered a regular policy, supported by the company's stable profitability and healthy operating cash flow.

Timothy Zhao from Goldman Sachs asked about Tuya's AI monetization strategy, including differences in pricing and cost compared to traditional IoT PaaS services. He also requested an outlook for Q2 and the full year on revenue and profitability, with a specific focus on the IoT PaaS segment's gross margin.

Answer

CFO Yi Yang stated that AI capabilities are seamlessly integrated into Tuya's existing PaaS, SaaS, and solution models, which allows for repricing certain offerings, such as AI-empowered audio PaaS. On the outlook, Yang acknowledged that tariff discussions have created short-term uncertainty and a 'bumping' in the demand cycle, but he expects a return to normal once negotiations are settled. For PaaS gross margin, he expressed satisfaction with the current high level, stating the focus is on maintaining it to motivate the ecosystem rather than pushing it significantly higher.

Timothy Zhao asked about the impact of AI on Tuya's smart solutions business, specific usage scenarios, and the outlook for margins and operating leverage in 2025.

Answer

Co-Founder and CFO Alex Young (Yi Yang) explained that initial AI opportunities are in audio/video interaction devices, complex decision-making scenarios like energy management, and AI-integrated toys. He stated that the company is prioritizing scalability over immediate margin impact for AI features. Young also clarified that OpEx is not expected to increase significantly as AI investment has been ongoing, and the company will maintain its focus on operational leverage.

Timothy Zhao from Goldman Sachs asked for color on the downstream demand outlook for the IoT PaaS business, including any differences across vertical segments, and inquired about the company's future shareholder return policy following the new anchor investment.

Answer

CFO Alex Yang described the current year as a 'rebounding year' for IoT PaaS as customers emerge from inventory cycles and reinvest in IoT projects. She highlighted two growth drivers: product innovation, particularly with AI-based services, and customer expansion into the enterprise level with solutions for telecom carriers and government projects. Regarding shareholder returns, she confirmed that the recent dividend was a special one, and future regular dividends will be considered annually based on non-GAAP profitability.

Timothy Zhao of Goldman Sachs asked about the outlook for operating profitability following the company's first profitable quarter, and also inquired about the long-term perspective on cash dividends.

Answer

Outgoing CFO Jessie Liu addressed profitability, stating the company expects to maintain a non-GAAP quarterly operating margin similar to Q2's 10% level, driven by the business model's operating leverage and disciplined cost control. Incoming CFO Alex Yang explained the dividend is a 'special dividend' based on H1 non-GAAP net profit, reflecting the company's strong cash flow and financial position, with the goal of establishing a sustainable, long-term shareholder return model.

Timothy Zhao asked about customer preferences and demand variations across VNET's wholesale campuses in different geographical locations like Hebei and Jiangsu, and sought insights into the pricing trends for the wholesale business for Q4 and next year, noting Q3 fluctuations.

Answer

Rotating President Ju Ma (translated by CFO Qiyu Wang) explained that customer preferences are driven by business type, proximity to headquarters, and ease of existing capacity expansion, leading to varying order paces across regions. CFO Qiyu Wang clarified that wholesale IDC pricing remained stable in Q3 2025, with faster-than-expected customer move-ins improving project IRRs. He emphasized VNET avoids low-price bidding in tight supply-demand areas.

Timothy Zhao from Goldman Sachs asked about customer preferences and demand differences across VNET's wholesale campuses in various geographical locations like Hebei and Jiangsu. He also sought clarification on the wholesale business pricing trend for Q4 and next year, noting some Q3 fluctuations.

Answer

CFO Qiyu Wang (and Ju Ma in the Chinese response) explained that customer preferences are based on business type, proximity to headquarters, and convenience for scaling existing capacity, with demand varying quarterly across regions like Greater Beijing and Yangtze River Delta. For pricing, Qiyu Wang stated that Q3 wholesale IDC pricing was fairly stable, with project IR often exceeding expectations due to faster customer move-ins. He emphasized that VNET avoids low-price strategies in areas with tight supply-demand dynamics.

Timothy Zhao of Goldman Sachs asked about the seemingly conservative second-half outlook implied by the full-year guidance and the reason for the sequential revenue decline in the retail IDC business.

Answer

VNet Management acknowledged the guidance is conservative, stating they are monitoring customer utilization pace relative to chip availability and may upgrade guidance if wholesale utilization remains strong. For the retail IDC business, they noted the slight sequential revenue decline was within a reasonable range and expect the segment to remain stable or grow slightly going forward.

Timothy Zhao from Goldman Sachs inquired about downstream customer preferences, asking if the AI boom has changed the balance between choosing in-house data centers versus third-party operators. He also asked how VNET's wholesale campuses cater to AI inference versus training demands and what specific customer requirements, such as power and latency, are for each.

Answer

An executive explained that while both self-built and rental models coexist, the vast majority of clients currently prefer renting from third-party operators like VNET due to the booming demand, making it impractical for most to build their own facilities. The prevailing model is 'rental plus self-build.' Regarding workloads, he clarified that AI inference models are typically deployed in retail IDCs, where clients prioritize convenience and application development over cost. In these scenarios, latency is not considered a significant issue.

Timothy Zhao pointed out a significant sequential increase in wholesale IDC monthly recurring revenue (MRR) and asked for the drivers. He also requested an outlook for the wholesale business's in-service capacity for Q4 and the following year.

Answer

CFO Qiyu Wang attributed the strong MRR growth to faster-than-expected client move-ins and improved profitability from economies of scale and lower costs. Rotating President Chenggang Shen reiterated that the delivery plan for Q4 remains on track to meet annual targets, and next year's capacity delivery guidance is based on secured orders and expected expansions.

Timothy Zhao inquired about the distinct pricing methodologies for AI training versus inference services at Kingsoft Cloud. He also asked about recent pricing trends for AI services over the past few months or year-to-date, and the differences in GPU utilization rates, pricing, and profitability between training and inference.

Answer

Zou Tao, Vice Chairman and CEO, explained that the pricing strategy for inference and training is largely similar, primarily based on resource usage. He differentiated between two inference models: customers buying resources and using the platform (similar margins to training) and customers directly purchasing API token services (expected to yield better margins). He noted that the API token services business is still in its nascent stages, and more time is needed to observe significant differences.

Timothy Zhao from Goldman Sachs inquired about the pricing methodology differences between AI training and inference demands, the pricing trends observed over the past few months or year-to-date, and the gap in overall utilization rates of chips/GPUs, pricing, and profitability between training and inference services.

Answer

CEO Zou Tao explained that the pricing strategy for inference and training is largely similar, based on quality and resources utilized. He differentiated between two types of inference services: customers buying resources and using the platform (with margins similar to training), and customers directly purchasing API token services, which are expected to yield a better margin ratio. He noted that the API token business is nascent and requires more time to observe significant differences.

Timothy Zhao from Goldman Sachs asked about the margin profile of the AI business and revenue from the Xiaomi/Kingsoft ecosystem compared to the overall public cloud, and also questioned the CapEx outlook and funding strategy for AI growth.

Answer

CFO Mr. Henry He stated that AI-related products deliver significantly higher gross and EBITDA margins than traditional cloud services. Regarding funding, he noted that the company has secured resources from banks and financial leasing companies and is applying for a new three-year related-party transaction cap of RMB 11.3 billion with Xiaomi and Kingsoft, ensuring strong shareholder support for CapEx without diluting public shareholders.

Timothy Zhao from Goldman Sachs inquired about Kanzhun Limited's progress in exploring different recruitment verticals and the potential long-term impact on services and monetization. He also sought management's view on the competitive landscape, particularly with general AI companies like OpenAI and Merkur entering the recruitment industry.

Answer

Jonathan Peng Zhao, Founder, Chairman, and CEO of Kanzhun Limited, shared insights from AI experiments, noting user behavior differences when interacting with AI versus humans. He emphasized a cautious approach to applying AI in recruitment, actively trying out new services like full-cycle or semi-cycle hosted recruitment in blue-collar and white-collar segments, but not yet ready for massive rollout. He acknowledged the interest of leading tech companies in recruitment but believes the key bottleneck is high-quality data, where Kanzhun and other industry peers hold an advantage, citing Merkur's focus on professional tagging.

Timothy Zhao from Goldman Sachs asked for updates on Kanzhun's exploration into different recruitment industry verticals and their potential long-term impact on services and monetization. He also sought management's view on the competitive landscape between traditional recruitment platforms, Kanzhun, and general AI companies like OpenAI and Merkur entering the recruitment industry.

Answer

Jonathan Peng Zhao, Founder, Chairman, and CEO of Kanzhun Limited, shared insights from AI experiments, highlighting the need for caution in applying AI to the 'ancient' job matching industry. He noted that while Kanzhun is actively but cautiously exploring hosted recruitment services in various blue-collar and white-collar segments with some initial achievements, a 'killer application' of large language models in recruitment has not yet emerged. Zhao acknowledged the entry of leading AI companies like OpenAI and startups like Merkur into the recruitment space, but asserted that the key bottleneck is high-quality data, where Kanzhun and its industry peers hold a significant advantage.

Timothy Zhao of Goldman Sachs Group, Inc. requested details on AI applications for enterprise clients and their commercial potential, and also asked about the company's capital markets strategy, including plans for Hong Kong liquidity and shareholder returns.

Answer

Wenbei Wang, Head of Investor Relations, described new AI tools like a recruitment agent named "HAMR" and customizable digital avatars for interviews, noting the company is patient for a "killer application" to emerge. CFO Yu Zhang explained the Hong Kong offering was aimed at improving liquidity, not raising capital. He highlighted the new $250M share repurchase program and an $80M dividend policy as strong commitments to shareholder returns, also noting the successful reduction in share-based compensation expenses.

Timothy Zhao asked for feedback on the company's internal AI feature testing, plans for AI monetization, and management's confidence in achieving the RMB 3 billion non-GAAP operating profit target for the year, including levers for H2 performance and capital allocation plans.

Answer

CEO Peng Zhao reported positive feedback from AI product tests, noting a 25% efficiency gain for recruiters and a 15% higher achievement rate for job seekers, but stressed a cautious approach to monetization. CFO Yu Zhang affirmed confidence in the RMB 3 billion profit target, citing disciplined cost control and operational leverage from internal AI. Zhang also confirmed the ongoing share repurchase program and stated the company is exploring other measures to enhance shareholder returns.

Timothy Zhao from Goldman Sachs asked for more detailed insights into the year-to-date recruitment demand trends across different enterprise sizes and industries. He also questioned the sustainability of the demand outlook and inquired about the company's strategy for its paying ratio and average revenue per user (ARPU) trends for the upcoming year.

Answer

Founder, Chairman and CEO Peng Zhao explained that the spring recruitment season showed strong sequential growth of 173% in active job postings, compared to 141% last year. He noted that the white-collar sector is stably recovering, boosted by AI-related job growth, and that overall cash-in is showing an accelerating trend. Regarding monetization, Mr. Zhao stated he has no aggressive plans to increase ARPU, preferring to support the market's recovery. He emphasized the company's potential lies in its ability to serve a growing number of new enterprises, having added over 6 million in the past two years.

Timothy Zhao requested a detailed breakdown of Q3 customer performance across blue-collar and white-collar sectors, as well as by enterprise size (SMEs vs. key accounts). He also asked about the remaining growth runway for individual users and the outlook for marketing expenses in the coming year.

Answer

CEO Peng Zhao noted that blue-collar revenue growth outpaced white-collar, now exceeding 38% of total revenue, with manufacturing and logistics as top-performing subsectors. He added that hyperscale (>10k employees) and small/micro (<100 employees) enterprises showed the strongest performance. CFO Yu Zhang specified that key accounts drove the highest revenue growth at over 30% YoY. Regarding user growth, Mr. Zhao stated the addressable market is over 400 million employees, providing a long runway for expansion, and confirmed that marketing spend will remain at a relatively low level next year due to strong organic traffic.

Timothy Zhao from Goldman Sachs asked for further details on KE Holdings' cost and expense control measures, the outcomes observed so far, and expectations for these lines going forward.

Answer

Tao Xu, CFO of KE Holdings, detailed cost reduction achievements across all business lines, including a 20% decline in fixed labor costs for existing home services, over 40% reduction in new home fixed labor costs, and over 20% decrease in procurement unit prices for renovation materials. He also noted optimized G&A and sales & marketing expenses, while R&D investments, particularly in AI, increased.

Timothy Zhao inquired about KE Holdings' cost and expense control measures, the outcomes observed so far, and future expectations for these financial lines.

Answer

CFO Tao Xu outlined optimization measures across all business lines: existing home services saw labor cost declines, new home services streamlined fixed and variable costs, home renovation reduced material costs through supply chain integration, and home rental services improved efficiency via technology and model refinement. He also noted G&A optimization, reduced sales and marketing spending, and increased R&D investment, particularly in AI.

Timothy Zhao from Goldman Sachs inquired about the performance of the secondary home market in the second quarter, its expected trajectory for the second half of 2025, and what potential policy tools might be implemented.

Answer

Executive Director and CFO Tao Xu explained that after a strong Q1, the market weakened significantly in Q2 with intensified divergence. He noted that while existing home transactions held up better than new homes, momentum slowed. Xu pointed to weak market sentiment and stated that future recovery depends on stronger policy support to restore confidence, with potential measures on both the demand and supply sides.

Timothy Zhao of Goldman Sachs Group, Inc. inquired about the forward-looking outlook for the Chinese property market, considering the post-Chinese New Year rebound, current macroeconomic dynamics, and the potential impact of U.S. tariffs.

Answer

Tao Xu, Executive Director and CFO, responded that while the existing home market saw a strong Q1 recovery, a typical seasonal slowdown is expected in Q2. Mr. Xu noted that external trade frictions have caused some short-term disruption in buyer expectations in certain cities, but overall homeowner sentiment remains stable. He expressed cautious optimism, believing that domestic supportive policies will likely mitigate trade risks and help stabilize the market in the medium to long term.

Timothy Zhao asked for management's outlook on the sustainability of the property market's recovery, particularly regarding year-to-date trends, and whether the company holds a more optimistic view for the upcoming year.

Answer

Tao Xu, Executive Director and CFO, explained that the market saw a strong rebound in Q4 2024 due to policy stimulus, with a notable structural shift toward existing homes. He reported that this momentum continued post-Chinese New Year, with existing home transaction volume up approximately 40% year-over-year. For 2025, Xu projects the market may begin to bottom out, contingent on price stabilization, forecasting a modest recovery for existing homes while the new home market may continue to adjust.

Timothy Zhao asked about the non-home transaction businesses, seeking details on the key operating focus for home renovation in 2024 and the outlook for the home rental business's economics and profitability drivers.

Answer

CEO Yongdong Peng stated that the home renovation focus has shifted from rapid scaling to building foundational capabilities, primarily by upgrading the Home SaaS system to version 2.5 and improving construction delivery. For the rental business, he reported that the number of managed units under the 'carefree rent' model surpassed 300,000. Profitability is being driven by improved unit economics through better leasing efficiency, reduced vacancy costs, and optimized re-renting processes. Peng proudly noted that the Beijing rental business achieved breakeven in H1 2024, validating the model's viability.

Timothy Zhao asked for details on KE Holdings' investment focus for its core home transaction business, including ROI evaluation and the latest updates on its connected store expansion. He also inquired about the outlook for new home sales GTV following a weak first quarter.

Answer

CEO Yongdong Peng stated the focus is on quality growth, with the number of active stores increasing 1.4% quarter-over-quarter and new stores showing high retention and productivity. He noted stores connected in Q4 2023 achieved positive ROI by March 2024. Regarding the new home outlook, Peng acknowledged market toughness but highlighted resilience through expanded developer partnerships and strict risk controls, evidenced by a 69-day DSO and a high percentage of commissions from SOE developers.

Timothy Zhao asked about the K-12 non-academic initiatives, specifically the enrollment growth gap between non-academic tutoring and the intelligent learning system and devices, and whether this gap is expected to sustain given the attractiveness of the intelligent learning system.

Answer

Stephen Yang, Executive President and CFO, clarified that the question was about K-9 business growth, not specifically the gap between non-academic tutoring and intelligent learning systems. He stated that junior high school business growth is faster than primary school due to a lower base and significant product development efforts, expecting this trend to continue with overall K-9 new business revenue accelerating to roughly 20% top-line growth in Q2.

Timothy Zhao asked about the enrollment growth gap between non-academic tutoring and intelligent learning system/devices within the K-12 non-academic initiatives, questioning if this gap could model revenue growth and if it's sustainable given the attractiveness of intelligent learning systems.

Answer

Executive President and CFO Stephen Yang, while addressing K-9 business growth, stated that junior high school business revenue growth is slightly faster than primary school business due to a lower base and significant efforts in developing new middle school offerings. He anticipates that middle school business growth will continue to be higher than kids' business, with overall K-9 new business expecting approximately 20% top-line growth in Q2 and hopes for better performance in the second half of the year.

Timothy Zhao from Goldman Sachs Group Inc. asked for a quantification of the impact from recent cost control measures on margins and inquired about the capacity expansion plan for learning centers in fiscal 2026.

Answer

Stephen Zhihui Yang, Executive President & CFO, estimated that cost controls could contribute 100-150 basis points to full-year margin expansion. He reported an 8-9% net increase in learning centers in Q4 and stated the FY26 plan is to align expansion with revenue, targeting a 10-15% increase, primarily in the second half of the year and in high-performing cities.

Timothy Zhao of Goldman Sachs Group Inc. asked for a quantification of the impact from cost control measures on the recent margin increase and the remaining room for such controls. He also inquired about the number of new learning centers opened in Q4 and the capacity expansion plan for FY26.

Answer

Executive President & CFO Stephen Zhihui Yang estimated that cost controls could contribute 100-150 basis points to margin improvement in FY26. He reported an 8-9% net increase in learning centers in Q4 and stated that the FY26 expansion plan targets 10-15% growth, contingent on revenue performance and restricted to high-performing cities.

Timothy Zhao from Goldman Sachs asked about the sustainability of operating margin expansion into fiscal year 2026, the specific measures being taken to enhance efficiency, and the company's strategy for balancing revenue growth with profitability.

Answer

CFO Zhihui Yang stated that he expects margin expansion in Q4 and fiscal 2026, driven by cost control and efficiency enhancements initiated in the current quarter. He highlighted a key focus on improving the utilization of facilities to support this goal.

Timothy Zhao from Goldman Sachs asked for details on the non-academic tutoring business, specifically the 26% enrollment growth this quarter and the expected normalized growth rate. He also inquired about the ramp-up pace and breakeven time for new learning centers.

Answer

Sisi Zhao, an executive at New Oriental, explained that quarterly enrollment figures can fluctuate due to registration timing. She stated the company is comfortable with a full-year growth rate of roughly 40% or more for the new business segment, driven by volume increases and a 4-6% rise in average selling price (ASP).

Timothy Zhao of Goldman Sachs asked for a revenue growth breakdown of the K-12 new initiatives and an explanation for the slowdown in non-academic tutoring enrollment growth. He later followed up with questions about East Buy, specifically its inclusion in the core guidance and the reasons for its implied revenue decline.

Answer

Executive Sisi Zhao stated that the key K-9 new initiatives grew over 56% in Q1. Executive President and CFO Zhihui Yang explained the enrollment growth slowdown was a timing difference from an earlier summer enrollment window, with combined Q4/Q1 growth remaining strong. Regarding East Buy, Mr. Yang deferred detailed comments to its own earnings release, while Ms. Zhao clarified that guidance comparisons are 'apple-to-apple,' including East Buy's former education business in both the current and prior-year base.

Timothy Zhao from Goldman Sachs inquired about TAL's Peiyou business, asking for updates on the offline learning center expansion pace, the business outlook for the year, and the ramp-up performance of newly opened centers.

Answer

Alex Peng, President & CFO, explained that while learning center expansion remains a key growth driver, the focus is on increasing density in existing cities with a disciplined approach. He noted that the rapid growth pace of fiscal 2025 is not expected to continue, and year-over-year revenue growth for Peiyou will likely taper. Peng added that new centers take slightly longer to ramp up but ultimately reach healthy operational levels, with overall Peiyou retention remaining strong at around 80%.

Timothy Zhao asked for details on the profitability profile of the Learning Devices (Content Solutions) segment in Q4 and the full fiscal year, and questioned the strategy to improve its profitability in fiscal 2026.

Answer

President and CFO Alex Peng confirmed the Learning Device business reported an adjusted operating loss for Q4 and the full fiscal year 2025. He explained that the current priority is building long-term competitiveness. For fiscal 2026, the strategy involves enhancing device functionality with AI, expanding the product portfolio, improving go-to-market capabilities, and driving operational efficiency through economies of scale. Peng characterized this as a multiyear capability-building phase paving the way for future growth and profitability.

Timothy Zhao of Goldman Sachs requested more insight into the growth drivers for the xueersi.com online enrichment programs, asking if growth was primarily from SKU expansion, product innovation, or aggressive marketing and user acquisition.

Answer

Executive Jackson Ding clarified that xueersi.com is a strategic priority where the primary focus is on product quality and excellence. He noted that growth is driven by both product development—including refining existing products and adding new SKUs and formats—and effective user acquisition. Ding emphasized that the company is leveraging multiple marketing channels to connect with a broader audience and drive user conversion.

Timothy Zhao questioned the 60% year-on-year growth in sales and marketing expenses, asking about the competitive landscape for learning devices and whether the marketing cost per device has increased, along with the future trend.

Answer

Executive Zhuangzhuang Peng acknowledged that the entry of 'full stack' competitors (with hardware, software, AI, and content) has intensified the market, which he views as positive for customer awareness and the industry. He admitted this impacts sales and marketing costs in some channels but stated the company's focus is on leading with product competency and exploring efficient go-to-market capabilities. While costs have increased in certain channels, they remain stable in others, and the company closely monitors efficiency across all channels.

Timothy Zhao from Goldman Sachs asked about supply and demand dynamics in China's Tier 1 data center markets, including industry-wide utilization and pricing. He also requested color on DayOne's projected 250MW of new commitments for 2025, asking about customer types, demand drivers, and any chip availability risks in Southeast Asia.

Answer

CEO William Huang stated that the Tier 1 market is starting to rebalance, predicting a potential supply-demand turnaround in 6-12 months which could improve pricing. For DayOne, Huang explained that new orders are from a diversified mix of cloud and video companies. He noted the primary driver in Southeast Asia is high-performance CPU for cloud and video applications, not GPU-driven AI, thus chip policies are not expected to have a significant impact.

Timothy Zhao from Goldman Sachs asked for management's view and strategy regarding AI empowerment in education, its application across business segments, and the potential for operating efficiency gains.

Answer

CFO Nan Shen detailed that Gaotu's current AI strategy focuses on internal cost reduction and efficiency improvements rather than external product promotion. Key applications include assisting in course design, automating homework correction, providing AI-driven oral language practice, and improving internal R&D and customer service systems. Shen emphasized a conservative investment approach, aiming to perfect internal use cases before marketing AI-powered services to new students.