Skip to content
UniqueCapital

UNIQUE RESEARCH / ENGLISH ARTICLE

Unique Friend | Biyou Tech’s Zhou Ze’an: AI Entrepreneurship Starts with Small Things

Unique Research · Insight / Unique Friend Q&A · January 17, 2024 · Guest: Zhou Ze’an, founder & CEO, Biyou Tech (必优科技) / ChatPPT

Inventory type remains insight (Unique Friend series — not partner welcome).

Guest

Zhou Ze’an — founder & CEO, Biyou Tech; serial AI founder; creator of ChatPPT and Pocket Animation; former Kingsoft Office product director and WPS KVP. Joined Kingsoft Office in 2013 — animation engine, smart docs, Daoke (稻壳); shipped WPS Office 2013/2016/2019. Early GenAI app pioneer: founded Pocket Animation in 2015 (acquired by Kingsoft 2017; then China’s largest Office extension by users); helped incubate Kingsoft Smart Docs (WPS AI, formerly Modou AI). At Biyou built WCCG LM framework and Wernicke document-creation middle platform, landing vertical models Bole, Shanggu, and Tuhu.

Zhuhai Biyou Tech (est. Oct 2020) focuses on AIGC for vertical content creation — a leading AI document-creation provider; backed by Kingsoft Office and Baidu.

Q&A

1. Product and core scenes?
Core product ChatPPT is the Office-plugin edition of the UTalk Doc-AI-Agent — AI presentation creation driven by natural-language UI (LUI) and Biyou’s Tuhu framework (LLM + WernickeLMs). Launched March 4, 2023 — 12 days before Microsoft announced Copilot — with 370M+ docs generated. Users interact end-to-end by language: topics or multi-doc input, plus language-driven editing. Six modules live: AI generate / beautify / present / co-write / motion / share — 1,152 feature classes. Editions: personal, enterprise, industry; online + plugin for individuals; SaaS embed / API for enterprises; industry SKUs for specialized content.

2. R&D / GTM — 2023 challenge and win?
Building fuller UTalk standalone (invite around Chinese New Year). 2M+ active users, 100+ enterprise customers, 85+ countries/regions; short-video views >200M in 30 days; awards from Baidu, iFLYTEK and peers. Biggest challenge: foundation-model surge hit the edge Biyou had built over two years on smaller generative models — forced a rethink of tech position and service paradigm, and ultimately birthed the true Doc Agent. Biggest win: integrating domain service models with LLMs into what Zhou calls the world’s first document Agent — opening more product imagination.

3. Challenges landing AI apps in China?
(1) Traffic — most apps are C-side; traffic is saturated and held by majors; without channels, no users. (2) App competition — crowded; startups and majors pile on any “good” idea; copycats and messy ops leave quality uneven and good apps under-heard. (3) Algorithm filing — China AI regulation is ahead in completeness, but filing is slow; without a listing, risk is high. (4) Base-model choice — domestic models still trail GPT; apps ride the base — weak base means weak user fit.

4. Competing with giants?
Three views: (1) Start small — pick a niche and go deep; giants swing high; startups can own a small cut and stay under the lamp. (2) Persist hard — once a niche is found, unravel every thread; that is how you share the table under giants’ noses — and maybe become a partner with their endorsement. (3) Speed — vs. giants “All In AI,” move 1–2× faster: fast trial, fast validation, fast response to tech and failed feedback. Keep running to earn a seat in the wave.

5. Are LLMs still too weak?
Disagrees that weakness directly blocks app progress. Limits are real and normal for emerging tech. Products must adapt to imperfect dependencies — imperfect LLMs still admit diverse solutions. If you wait for the base to be fully ready, why wouldn’t the model vendor solve the user problem instead of you? Products are staged (MVP → PMF → UP); the model is a base that can have interim substitutes. Adapt to user, product, and market — don’t wait.

6. 2024 tech to watch?
(1) Agents — best practice for vertical LLM apps and the best transitional state toward AGI; OpenAI, Baidu and vertical practitioners all investing. (2) Multimodal generation — image, speech, video, 3D unlocking generative content value faster post-LM boom. (3) Text-to-video — short-video era already proved video beats classic text/image for distribution; fast landing is both multimodal product and AI breakthrough.

1 CN image; attach on ingest. Signup CTA in CN omitted.

— END —

Editorial English edition by Unique Research, published on UniqueCapital on January 17, 2024. This translation is based on an earlier Chinese article; the Chinese source is credited separately and is not treated as a line-by-line mirror.

View the Chinese source article ↗
← Back to English research