Early this month, iFlyrec, a product under iFlytek, announced that it had surpassed 100 million users, becoming the second of the company’s core offerings to cross the nine-figure threshold after its input method solution.

At a time when many artificial intelligence startups are chasing scale through subsidies and traffic-driven growth, iFlyrec has taken a different path. It built its business on paid technology services and subscriptions, carving out a software-as-a-service model that runs counter to the fundamentals of traffic arbitrage in the internet era. Its rise tests the commercial viability of applying the SaaS model to AI solutions in China and offers a case study in how to grow without joining the subsidy race.

Amid rising customer acquisition costs and recurring subsidy wars, iFlyrec’s strategy appears contrarian. Since its launch, it has maintained that it would not rely on advertising. It abandoned ad-driven monetization from the outset and committed instead to a subscription-based SaaS model.

That choice appears to have produced measurable results. According to its 2025 financial report, the iFlyrec app has recorded gross profit growth of more than 60% for three consecutive years. Its annual renewal rate for new users exceeds 50%, and it consistently ranks among the top apps in major app stores in the speech-to-text category.

Behind those numbers is the logic of the SaaS model. Paid access filters for users with clear, recurring needs. Ongoing technical upgrades support retention. Together, they create a cycle of product iteration, paid conversion, and renewal. Unlike traffic-driven utility apps that depend on short-term user spikes, SaaS products rely on cumulative user value and predictable cash flow.

Competition in this segment has never been gentle, and the pandemic has only accelerated the commercialization of collaborative office SaaS platforms. Products such as Feishu (Lark), DingTalk, and Tencent Meeting embedded voice transcription features into their offerings, leveraging existing ecosystems to attract users. At the same time, hardware devices tailored to Chinese consumers’ willingness to pay for dedicated tools, including recording cards, entered the market. The result is a competitive structure that combines software integration with hardware participation.

But will users choose an all-in-one office suite for a single transcription feature?

Collaboration platforms prioritize teamwork, and voice transcription is only one of many auxiliary functions. That makes deep scenario optimization or technical specialization difficult. Hardware products often perform well in noise reduction and battery life, but they still rely on software to transcribe and organize content.

By focusing narrowly on speech transcription and AI-generated meeting summaries, iFlyrec has taken a more focused approach. It targets users who process large volumes of audio and avoids direct competition with collaboration platforms. In a crowded market, defending and refining a vertical niche may prove more sustainable than competing on breadth.

In the era of large language models, or LLMs, AI’s primary interface is shifting from typed text to natural interaction. Voice, as a low-friction form of input, is emerging as a potential gateway.

Unlike text input, which requires deliberate structuring, or image and video input, which often require specific devices and settings, voice can be generated in most environments. It does not require additional training and integrates into meetings, interviews, training sessions, and research. This ubiquity positions voice as an entry point through which AI can enter daily work and life.

iFlyrec’s second core advantage lies in its timing. Rather than building a comprehensive office platform, it concentrated on speech transcription and summarization while aligning with voice as a widely used entry point in the large model era. This positioning supports its ability to surpass 100 million users.

From 2015–2019, the product relied primarily on non-real-time transcription as its core business, building an initial user base. In 2019, it shifted toward subscription-based SaaS services, streamlined its product structure, expanded usage scenarios, and prioritized real-time recording and transcription. The transition aligned with the subscription logic of SaaS and coincided with the surge in LLMs beginning in 2023.

In May 2023, iFlyrec integrated iFlytek’s Spark LLM into its product, combining speech technologies with generative AI capabilities. It introduced a “single-click” workflow to address a common user complaint: transcripts that capture content but lack structure. The update marked a shift from pure transcription to structured organization.

As competition in LLMs intensifies, the product has evolved further. Its AI-driven summary feature now includes an adaptive summary mode. According to the company, the feature adjusts output based on users’ notes, inserted images, and highlighted sections, rather than relying solely on fixed templates.

Its ambitions extend beyond software.

iFlytek’s speech technologies have been deployed in several high-profile scenarios, including real-time transcription during the Miss Hong Kong Pageant live stream, real-time subtitles for the Spring Festival Gala produced by China Media Group, and an AI-powered real-time captioning system for the Hong Kong Legislative Council as part of a digital governance initiative in the Guangdong-Hong Kong-Macao Greater Bay Area.

Together, these cases illustrate a broader strategy: using SaaS products as a core offering while extending capabilities across software, hardware, and services to serve government, media, and enterprise clients. The approach reflects iFlytek’s broader business framework and suggests a path for vertical SaaS products seeking to expand commercial value without diluting their specialization.

Many AI products equate scale with success, sometimes at the expense of sustainable revenue. Yet competition is shifting from all-encompassing platforms to deeper vertical specialization. As enterprises advance digital transformation, their operational demands are becoming more granular, and broad platforms may struggle to address every niche requirement.

For AI SaaS developers, long-term viability depends on technological discipline, close engagement with user needs, and a durable revenue model. In an industry inclined to prioritize rapid scale, iFlyrec’s trajectory suggests that growth without heavy subsidy spending remains possible, provided the product can sustain recurring value.

KrASIA features translated and adapted content that was originally published by 36Kr. This article was written by Xiao Xi for 36Kr.