Original · Unique Research · 2026-07-12
Editor's note: This is a follow-up interview with Zhou Ze'an (周泽安), founder of Biyou Technology (必优科技) and creator of ChatPPT. All statements, user figures, and product details are attributed to the interviewee. Company and product names are preserved as stated.
AI Industry Observation
"There Will Be No More PPT in the AI Era"
Zhou Ze'an, a 13-Year Document Veteran: Don't Make a "Faster Tool," Make a "Better Result"
This is what Biyou Technology founder Zhou Ze'an told me in March this year—after all, Biyou holds 17 million ChatPPT users and 80 million AI document registered users, not just talk. By July 2026, I'm sitting across from him again. My first question: did that harsh prediction of yours come true? Zhou Ze'an smiled: "The direction is right, but the pace is slower than I expected." Hidden in this sentence is the truth of the entire AI office industry.
"Instant Presentation" Works, But the Mainstream Hasn't Arrived
What Zhou Ze'an calls "instant presentation" is essentially revolutionizing PPT. The traditional logic is: first write a Word document, then make it into a PPT, end-to-end transform it into a presentation version, and finally stand in front of the projector and talk. Three steps, three tools, three file formats. Zhou Ze'an's prediction: these three things will collapse into one—you state your need, and AI directly gives you an interactive, presentable, shareable "thing." Put simply, not "edit then present," but "present is edit, edit is present."
Technically, this has been achieved. ChatPPT's AI real-time roadshow feature, multi-source views, shared interactions—essentially all doing "presentation is delivery." You can also see the trend externally—now using Claude, Codex to directly write interactive presentation pages is no longer novel. The team is still polishing a more native "instant presentation" upgrade project internally, reportedly going live soon. But between technical achievement and mainstream adoption, there's a deep chasm.
Zhou Ze'an candidly says there are two bottlenecks. The first is technical and product maturity, especially requirements for foundation models—"AI delivering presentation, it can't not be instant; instant response but poor delivery quality is also no good." Speed and quality need to improve in the same direction; this isn't something one startup can solve alone. When he said this, there was no anxiety in his tone, only the patience of someone who's built document products for thirteen years. The second bottleneck is more interesting: user cognition. We've been educated by Office for thirty years. Thirty years of muscle memory can't be changed in twelve months. Zhou Ze'an described a very fine but representative phenomenon: users need a "file" to confirm "I'm done." Give them a link, and they'll feel "this isn't finished yet, right?" The reason PPTX, DOCX, PDF formats are stubborn isn't just technical inertia; it's psychological anchors—that save button, that file icon, that "downloaded locally" sense of certainty—constitute an entire generation's definition of "done."
"It's not that technical standards aren't ready; it's that user cognition standards haven't arrived."
Pay for Certainty, Not for Speed
After discussing the technical route, I raised a more practical question: how do you actually make money? Zhou Ze'an's answer is as crisp as a year ago—"result delivery." But over this past year-plus, his understanding of "results" has deepened. Users won't pay for "saved 10 minutes," but they will pay for "not having to revise it a third time." This sentence is worth unpacking. Most AI tools sell "speed"—you used to write in a day, now write in an hour, so I charge you a few dozen yuan a month. This logic has a problem in Zhou Ze'an's view: what users truly care about isn't time, but certainty. AI gives a "result that can be used directly"—not a half-finished product, not a draft needing further processing, but a finished product that can be sent directly to clients, to leadership, used in formal settings. This is the underlying motivation for payment.
What surprised me more is that over the past year, a paid feature emerged that Zhou Ze'an himself hadn't fully anticipated: trusted traceability service. When ChatPPT generates every page of content, it can trace back to "which paragraphs from which articles this viewpoint comes from." Whether it's web-scraped data or user-uploaded documents, every piece of content has source annotations. Zhou Ze'an even led his team in doing something very "geeky"—launching an AIPC local traceability version covering Intel, AMD, and Snapdragon chip platforms, so traceability can be done locally, not leaving the device, not going to the cloud, secure and trustworthy. Why do office scenarios especially need this? Because PPTs in the workplace aren't like posting to Moments—every data point, every viewpoint you write on a page may have an accountability hearing behind it.
"The embarrassment of being asked speechless in a conference room costs far more than a few dozen yuan a month subscription."
Trust and Thinking: Two Real Problems of AI Office
In last year's interview, Zhou Ze'an raised two concepts: "content trust problem" and "self-thinking problem." Over a year later, the first has a solution, the second is still being explored. The solution to content trust is traceability. ChatPPT now achieves that when generating every page of content, it can tell you "which paragraph, which Q&A this viewpoint comes from." Auditable version records are also available in the enterprise version. Zhou Ze'an's team's thinking is clear: when AI writes things for you, you must know where it learned from—it's not an optional feature, it's the bottom line.
But what's more interesting is another still-brewing feature—"Echo Mode." Zhou Ze'an explains that "anti-optimization" isn't deliberately speaking poorly, but not helping the user smooth things too much. AI polishing has a natural tendency: turning your colloquial expressions into formal prose, your fragmented sentences into parallelism, your "I think" into "research shows." It looks "more professional," but it also polishes away your original self. The idea behind "Echo Mode" is: preserve the user's original expression's tone, phrasing, even colloquial habits. You're not writing a "standard answer"; you're expressing yourself. AI doesn't speak more beautifully for you; AI helps you say what you want to say more clearly.
And the highest form of "result delivery," in Zhou Ze'an's view, is AI roadshow. Previously, "the result" was a PPTX file. Now? It can be a complete roadshow package with AI script and voice cloning, an online interactive presentation link, or a batch of content already adapted to enterprise brand guidelines. ChatPPT's AI roadshow feature can automatically generate scripts, clone voices, even produce playable roadshow videos. A user uploads a business plan; AI doesn't just make it into a PPT—it can help prepare the script, clone your voice for rehearsal, generate a roadshow video that can be sent directly to investors.
"
Users don't just want it "made"; they want something they can take and use.
What Big Tech Disdains, Users Can't Live Without
Kingsoft and Baidu are both shareholders of Biyou Technology. Zhou Ze'an didn't hedge his answer: "Overall tighter, but tension is inevitable." On the Kingsoft side, Zhou Ze'an's team's very origin is a trust card—a team that came out of Kingsoft naturally has low communication cost. But the functional overlap between WPS AI and ChatPPT is obvious. Zhou Ze'an's strategy is clear:错位. WPS AI is strong in "ecosystem and comprehensive native"—write documents, make spreadsheets, do presentations, all in one. ChatPPT is strong in "vertical presentation scenarios + full-link native"—from writing content to making presentations to roadshow, all in one product. On the Baidu side, collaboration is deeper. Many native document components in Wenxin Qianfan have Biyou Technology's underlying technology output; more importantly—Baidu's Wenxin Assistant and search's AIPPT creation feature are all exclusively powered by ChatPPT's technology service. This means Biyou Technology isn't a "portfolio company" of big tech; it's a "necessary component" in big tech's ecosystem.
But this position isn't permanent. Zhou Ze'an is blunt: the opportunity window is narrowing. Doubao, Kimi, WPS AI, iFlytek Zhiwen—all iterating rapidly. Big tech has money, people, compute; why do startups survive? His answer comes back to that keyword—"scenario depth." "Preserve original text, traceable generation"—put in front of a big tech product manager, this doesn't rank in the top fifty priorities. Big tech does "general capability"; to serve 100 million people, you can't specially optimize a feature for just finance, law, and academia. But this very feature is lifeline for ChatPPT's core users—you write a prospectus at an investment bank, every data point needs a source; you write a due diligence report at a law firm, every viewpoint needs a basis. Traceability isn't a value-add; it's an industry admission threshold. Big tech does "general capability," startups do "scenario depth"—this division still holds, but the window is narrowing. Zhou Ze'an's judgment is that this window may only have twelve to eighteen months left.
Defining the Format Isn't Writing a White Paper; It's Getting Users to Actually Use It
At the end of last year's interview, I asked Zhou Ze'an what he most wanted to accomplish next. He said three things. A year later, the answer hasn't changed, but the pressure has grown. First: turn "instant presentation" from concept into a scalable product. Second: complete the Beta version definition of the "AI-era document format." Third: take at least 15% of the 500 enterprise clients to complete the upgrade transition to the new document format. Of these three, the first is the life-or-death line. "If 'instant presentation' can't form a scalable user base in the next 12 months, then the judgment I made in 2025 needs to be re-examined—not that the direction is wrong, but that the team's execution has a problem."
Speaking of which, I noticed a detail: Zhou Ze'an is still personally overseeing the product. Not "inquiring," but real frontline decision-making. "I've always been the company's biggest PM." When he said this there wasn't a trace of pride, like stating a fact. Previously the front line was writing code, drawing prototypes; now he uses Vibe Coding to work out approaches and methods, then clashes with the product R&D team, and the front line is also meeting clients, looking at data, watching competitors.
"Don't make a 'faster tool,' make a 'better result.' When the entire industry is competing on who generates ten seconds faster, Zhou Ze'an and his 80 million users are actually waiting for something slower and harder to happen—not that tools got faster, but that 'the result' itself changed meaning."
Interview Selected Q&A
Q1｜In your March 2025 exclusive interview you said something harsh: "There will be no more PPT in the AI era; future documents will be instant presentation." Over a year later, has this judgment been validated?
Zhou Ze'an: Honestly, the direction of this judgment is right, but the pace is slower than I expected. "Instant presentation" is technically already achieved. Take ChatPPT as an example, AI real-time roadshow, multi-source views, shared interaction modules are essentially doing "presentation is delivery." Our team is also currently focusing on polishing a more native "instant presentation" upgrade project, expected to meet everyone soon. Also, from the external environment, market acceptance of instant presentation is improving. At least now everyone has started using Codex, Claude to directly write pages with interaction and delivery results, which shows that "documents aren't necessarily just static files" is already understood by more and more people. But it hasn't become mainstream; there are two core bottlenecks. First, technical and product maturity is still being validated. Second, users also need to adapt to "delivery-result-level products."
Q2｜You said at the time that what you most wanted to do in 2026 was "define a document format unique to the AI era." Now it's July 2026—has this "new format" been defined?
Zhou Ze'an: Not completely defined, or not officially launched yet, but the prototype exists. Internally we call it "LiveDoc's native document format." It's not a static file, but an "interactive, feedback-capable content container." Its fundamental difference from PPT, Word, Excel is: traditional formats are "edit then present"; what we hope to do is "present is edit, edit is present." Currently this new format hasn't truly been delivered to users, but it will be soon. Its biggest obstacle isn't technical standards, but "user cognition standards." Users' understanding of "documents" has been educated by Office for thirty years; they need a "file" to confirm "I'm done." Looking back now, the expectation of completing definition in 2026 was indeed slightly optimistic. Official launch may be later in the second half of the year.
Q3｜You emphasized at the time that users pay not for efficiency tools, but for "pre-rendered results" and "result delivery." Over a year later, does this payment logic still hold?
Zhou Ze'an: It still holds, and is increasingly validated. The core reason users are willing to pay is that AI gives them a "result they can use directly." Not a half-finished product, not a draft needing extensive further processing, but a finished product that can be sent directly to clients, to leadership, used in formal settings. This logic hasn't changed. Instead, as AI PPT generation technology becomes more widespread, users' requirements for "result quality" have gotten higher. At the same time, new paid points have emerged, like trusted traceability service. We can now support viewpoint-level traceability—that is, generated content can be traced back to which articles, which paragraphs. To support this, we also launched AIPC versions including Intel, AMD, and Snapdragon chips, supporting local-level traceability.
Q4｜You mentioned two new AI office problems before: content trust and self-thinking. Do these two problems have solutions now?
Zhou Ze'an: The content trust problem has a solution; the self-thinking problem is still being explored. On content trust, the traceability mechanism is already live. When ChatPPT generates every page of content, it can tell you: this viewpoint on this page comes from which paragraph, which Q&A. Auditable version records are also already done in the enterprise version. As for self-thinking, internally we've discussed the "anti-optimization" feature many times but haven't officially launched it. Our approach isn't to deliberately speak poorly, but to add an "Echo Mode"—preserving the user's original expression's tone, phrasing, even colloquial expression habits, without over-polishing.
Q5｜After the Agent concept exploded, many products changed from "generation tools" to "execution assistants." Has ChatPPT's positioning changed?
Zhou Ze'an: The positioning hasn't changed, but the definition of "result" is changing. Previously, "the result" was a PPTX file. Now, "the result" can be a complete roadshow package with AI script and voice cloning, an online interactive presentation link, or a batch of content already adapted to enterprise brand guidelines. So "result delivery" as the core hasn't changed, but the forms of results are expanding. ChatPPT's AI roadshow feature can automatically generate scripts, voice cloning, even produce playable roadshow videos. Users don't just want it "made"; they want something they can take and use.
Q6｜Over a year later, have users' understanding of AI office matured? Or because AI is too widespread, are expectations for "one-sentence generation" even higher?
Zhou Ze'an: Polarized. Part of users, especially enterprise-level users and high-frequency users, have very mature cognition. They know AI generation is just the starting point; the subsequent refinement, adjustment, adaptation is the bulk. But another part of users, especially C-end low-frequency users, because AI is too widespread, have even higher expectations for "one sentence generates a perfect PPT."
Q7｜You previously positioned yourself as a "true product practitioner and frontline explorer in the AI office wave." Now the company size, product matrix, and user scale are larger—do you still feel you're on the front line?
Zhou Ze'an: I must be on the front line, and it's "multiple front lines." Previously my front line was writing code and drawing prototypes. Now I still use some Vibe Coding approaches to clash with the product R&D team. But today's front line isn't just writing code; it also includes meeting clients, looking at data, watching competitors. With AI enhancement, I have more "multi-line" energy to feel the market temperature. I still personally make product decisions now, and quite a few of them. I've always been the company's biggest PM.
Q8｜Kingsoft and Baidu are both your shareholders. Over a year later, is the "big-tech shareholder + startup" relationship tighter, or has tension emerged?
Zhou Ze'an: Overall it's tighter, but tension inevitably exists. On the Kingsoft side, we're a team that came out of Kingsoft, so communication cost is low. WPS AI and ChatPPT do have functional overlap, but our strategy is错位 competition. WPS AI is strong in "ecosystem and comprehensive native app convenience"; ChatPPT is strong in "vertical presentation scenarios + full-link native." On the Baidu side, collaboration is also very tight, with technical-level synergy. Many native document components in Qianfan have Biyou Technology's value. At the product level, the AI PPT creation parts in Baidu's Wenxin Assistant and search are all exclusively powered by ChatPPT's technical service.
Q9｜You said at the time that foundation-model giants do general, startups do scenario insight, so for Biyou Technology opportunity outweighs challenge. Now big tech is investing more in office scenarios—does this judgment still hold?
Zhou Ze'an: It still holds, but the opportunity window is narrowing. Big tech is indeed increasing investment; Doubao, Kimi, WPS AI, iFlytek Zhiwen are all iterating rapidly. But big tech does "general capability," startups do "scenario depth." For example, features like "preserve original text, traceable generation"—big tech won't specially optimize for just finance, law, and academia vertical scenarios. But for our core users, these features are lifelines.
Q10｜If you could only All in on one direction, would you continue deep-diving "document result delivery," or expand horizontally to the full office workflow, or go to the infrastructure layer?
Zhou Ze'an: All in "document result delivery." It's not that I don't want to do horizontal, but that I can't. Biyou Technology's core capability is in the "document" vertical. We have 13 years of document accumulation, 30 vertical models, 500+ enterprise clients. Horizontal expansion means re-accumulating for every scenario, and we don't have big tech's resource depth. Some people will question whether deep-diving documents means a low ceiling. But I think this is a false proposition. The global intelligent document market is over $100 billion; just original intelligent PPT alone is over $4 billion. Our strategy is "use PPT as the entry point, penetrate into the full document scenario."
Q11｜If you could start over, when founding Biyou Technology in September 2020, what decision would you do differently?
Zhou Ze'an: Technical route. In 2020 we took the "small model + vertical scenario" route, starting with specific scenarios like resumes and contracts. This choice was right at the time; it helped us accumulate complex details in real document scenarios. But if I could have foreseen the speed of large model explosion in 2023, I would have shifted resources earlier toward the "large model + vertical model" dual-engine drive.
Q12｜If you could only give one piece of advice to attendees of the Unique Awards about "how to survive in AI office," what would you say?
Zhou Ze'an: Don't make a "faster tool," make a "better result." Users won't pay for "saved 10 minutes," but they will pay for "not having to revise it a third time." The ultimate value of AI office isn't "efficiency," it's "delivery."
Q13｜As a "13-year document veteran," what are the three things you most want to accomplish in the next 12 months?
Zhou Ze'an: First, turn "instant presentation" from concept into a scalable product. Second, complete the Beta version definition of the "AI-era document format." It doesn't have to become a standard this year, but at least there should be a usable product prototype and first batch of seed users. Third, take at least 15% of the 500 enterprise clients to complete the new document format upgrade transition, and complete final delivery, improving industry replicability. Of these three, the most critical is the first. If "instant presentation" can't form a scalable user base in the next 12 months, then the judgment I made in 2025 needs to be re-examined. Not that the direction is wrong, but that the team's execution has a problem. "Defining the document format" isn't writing a white paper; it's getting users to actually use it.