With listed companiesWith the arrival of the 2025 annual report disclosure season, people have been able to glimpse the catalytic effect of AI on listed companies' performance. Among them, hardware manufacturers such as Industrial Foxconn, Inspur XCUOCHENG, and Cambricon, which are the "scraping contractors" in the upstream AI industry chain, have seen a surge in performance that has confirmed the high industry prosperity brought about by the explosive demand for AI. However, to verify the progress of AI commercialization, attention needs to be paid to the midstream platform layer and downstream application layer. From the listed companies' 2025 performance, has the commercialization of AI applications been verified? Has the shift from "speculating on concepts" to "looking at performance" been completed? Which types of companies (such as cloud providers/vertical SaaS) have achieved the most successful commercialization of AI? What is your most concerned indicator, such as the proportion of AI revenue or changes in gross profit margin?
Frost & Sullivanfrost &Lu Jing, China Business Partner and Managing Director of Frost & Sullivan (hereinafter referred to as 'Frost & Sullivan'), was interviewed by the Financial Times to discuss the commercial validation of AI.
Q: It is generally believed2025The year isAIThe first year of application. Listed company2025In the annual results,AIHas the commercialization of applications been validated? Has the shift from 'speculating on concepts' to 'evaluating performance' been completed?
I thinkAIApplication commercialization in2025The year has been partially validated, but it has not yet entered the full performance realization phase. From the financial report, what the market is most clearly seeing are two types of changes: First,AIWe have started to substantially drive revenue growth rather than merely boosting market sentiment. Second, some companies have seen synchronized improvements in profit margins, cash flow, and subscription metrics, which meansAIIt's not just about feature upgrades, but about reconstructing the business model.
Take US stocks as an example.AIThe commercialization of applications is first realized in high-frequency workflows and enterprise productivity platforms.Adobe,ServiceNow,SalesforceWaiting for the company'sAIThe related products have all achieved significant growth. These data indicate that the global market is no longer just about having itAIStory, rather than just starting to readAIWhether it can drive subscription, contract volume, renewal, and enterprise spending expansion.
In addition, the industry is still in a clearly differentiated phase, with some companies having provenAICan bring new revenue and profit leverage, while some companies are still stuck at feature display,POCVerification or pay-as-you-go for low-permeability scenarios.2025The annual financial report has been verifiedAIThe application is not a false positive, but not all connections have been verified.AICompanies in the industry can all successfully deliver their performance. From an investment perspective, the market has indeed moved on from whetherAIAbility switched toAIWhether the income structure, profit margin, and cash flow can be changed, but this transition is still ongoing and has not been fully realized.
Q:From a financial report perspective, which types of companies (such as cloud providers/verticalSaaS)AIIs commercialization the most successful? What are the most concerned metrics, such asAIChanges in revenue proportion and gross profit margin?
from2025Looking at the year and the most recent full financial reporting cycle,AIThe most commercially successful companies are mainly concentrated in three categories. The first category is consumer-oriented high-frequency tool companies, which have clear user scenarios, extremely short paid conversion paths, and subscription-driven business models.AIThe most direct manifestation is user growth,ARPUEnhancement and profit release. The second category includes enterprise software and productivity platform companies, includingAdobe,ServiceNow,Salesforce,Zoom inThe common characteristic of these companies is that they are already located at the center of the customer's daily workflow.AIIt's not about creating demand from scratch, but about enhancing automation, reducing manual operations, and improving the customerROITo drive penetration of additional purchases, renewals, and higher-tier packages.ServiceNowandSalesforceEnterprise-levelAIThe most effective commercialization approach is toAIEmbedding into work order flow, customer service flow, sales flow, and enterprise data layers can translate into growth in contract amounts and remaining performance obligations. The third category is vertical industry solution companies, which often do not burst onto the profit statement as early on but are more likely to show early signs in terms of orders, collections, scenario penetration, and large customer signings.
As for the most important indicators, I don't think we can just focus onAIIncome proportion. Because many companies haven't yetAIIf income is separated out, this alone can underestimate the true progress. I am more concerned with five types of indicators: first,ARR,RPOContract liabilities, number of subscribed usersAIWhether it generates sustainable revenue; secondly, gross profit margin and operating profit margin, to be consideredAIWhether there is economies of scale, rather than just increased reasoning costs; thirdly, operating cash flow and sales receiptsAIWhether the revenue can really be recovered; fourthlyAIWhether the core business revenue growth rate of the driver is faster than the overall company growth rate; fifthly, the usage depth at the customer level, such as in large customer ordersAIpermeability,AI MAU,AIPackage add-ons, etc. In short, determineAIThe commercialization of an application doesn't depend on what models are released, but onAIHas the quality of the company's revenue and operating leverage changed?
Q:AchievedAIWhat commonalities do commercial companies possess? Data barriers, customer base, industryKnow-howTechnical capabilities — which is the most critical? Which segments are easier to break into and dominate?
I believe it can be successfully achievedAICommercial companies typically possess four common characteristics: First, they have high-frequency, rigid demand, and are quantifiable.ROIApplication scenarios; secondly, having a native customer base and mature distribution channels; thirdly, having data, process, and industry knowledge accumulation; fourthly, being able to encapsulate model capabilities into products and workflows, rather than staying at the technical demonstration level. I believe the most critical aspect is the industryKnow-howAdd workflow integration capabilities. As underlying model capabilities are accelerating their diffusion, simply calling the modelAPIIt's difficult to build a long-term moat; what really determines who can make money is who understands business processes better, who understands customer pain points better, and who canAIEmbed it into the existing software and service system, and transform capabilities into efficiency gains, price increases, or additional purchases.Adobe,ServiceNowandSalesforceThis is the commonality,AIAbility is not a cheat code, but part of the core capabilities of the platform.
From the perspective of the track, I think the one that is most likely to take the lead isROIThe track with the clearest vision, the shortest procurement decision-making process, and the deepest product integration. Currently, at least three tracks are more likely to lead in delivering value first. The first is content creation and personal productivity tools, including images, videos, design, office software, language learning, etc., because users use them frequently, the subscription model is mature,AIEnhancing user experience is intuitive and can be easily translated into paid services. The second category includes enterprise productivity and process software, such asCRMCollaborative Office, Customer Service, and Sales Automation, becauseAIIt can be directly translated into increased employee efficiency and reduced operating costs, making companies more willing to pay for it. Thirdly, there are vertical scenarios with strong industry process loops, such as industrial production, healthcare, education, etc. Although the implementation pace of these tracks may be slower than consumer subscriptions and standardization,SaaSHowever, once verified, customer stickiness and entry barriers are often higher. In contrast, those lacking clear scenarios, with unclear sources of customer budgets, and relying solely on one-time project revenues are moreAIApp companies, on the other hand, find it even more difficult to lead the way in financial reporting.
*This interview was published inFinancial TimesReporter: Li GuohuiOriginal title:Unveiling the Secrets of the Earnings Season:AIHave you started making money?
