Why Chatbots Capture Only 10% of the AI Market and Enterprise Agents Hold the Real Gold

At the 2026 Zhongguancun Forum, Kunlun Wanwei unveiled three world‑class AI models and a "3+1" AGI strategy, with CEO Zhou Yahui emphasizing that chatbots represent only one‑tenth of the largest market, while enterprise‑grade AI agents constitute the real growth opportunity.

AGI Defined as Automatable Workflow Capability

AGI can be understood as "the system capability to automate digital and physical workflows", not a sci‑fi superintelligence. By this definition, AGI already exists in certain domains.

Zhou argues that this pragmatic definition turns AGI from a philosophical question into an engineering problem that can be measured and tracked.

Three New Models – Video, Music, and Gaming

SkyReels V4 is an AI video‑generation model powering the DramaWave short‑film platform, offering "director‑level control" through text, image, and multimodal inputs. Zhou claims it currently ranks first globally in the video‑generation lane.

Mureka V9 is an AI music‑generation model behind the Mureka platform, positioned as an "AI‑powered Spotify". It also leads globally and is the most commercially mature product for Kunlun Wanwei.

Matrix‑Game 3.0 is a game‑world model supporting the Cat Forest Academy 2.0 platform, aiming to generate not only visuals but also physics and logical rules. Zhou notes that a GPT‑4‑level model for this domain may not appear until 2028.

Market Insight: Chatbots vs. Enterprise Agents

Chatbots account for only one‑tenth of the biggest market; the real opportunity lies in the enterprise software market for productivity agents.

Zhou observes that this view aligns with Silicon Valley sentiment that AI agents will reshape SaaS, displacing legacy products like DingTalk and Feishu. He likens Kunlun Wanwei’s strategy to “following ByteDance’s path but using domestic resources.”

The “One‑Person Company” Vision

AI can replace a finance manager, sales staff, and growth team, handling tasks for a solo entrepreneur.

Future agents could negotiate with customers on your behalf.

According to Zhou, this underpins the “Skywork” AI operating system aimed at creators and one‑person companies, where all data and technology target agents rather than humans.

Financial Outlook and Pricing

Maintaining a top‑tier model requires at least ¥100 million of compute per month. The company has lost ¥1.2‑1.5 billion annually for the past three years. Zhou projects rapid revenue growth in 2026, profitability from 2027‑2028, and confirmed profitability after 2028.

He cites Kimi’s recent surge—after releasing a 2.5‑billion‑parameter code model, revenue multiplied several‑fold within a month—as evidence that strong models drive income, provided they remain differentiated.

New models will be priced higher; token costs for programmers will rise, financial agents may command premium rates, while general models trend cheaper.

Sora’s Failure and Opportunities in Video Generation

Sora’s failure reflects a product misstep, not the viability of video generation.

Zhou attributes the setback to OpenAI’s strategic choice to focus on productivity agents over video, and to the loss of key video‑generation engineers to Meta.

He believes the gap creates an opening for Chinese companies, though competition remains intense.

Structural Challenge in China’s AI Industry

Most Chinese robotics firms wait for Musk’s Optimus before building their own models.

Technical staff lack decision‑making power, while decision‑makers lack technical understanding—a structural issue.

Zhou admits Kunlun Wanwei has experienced this mismatch and is now involving technical teams directly in product decisions, though the problem persists industry‑wide.

Experience links: SkyReels V4 – https://www.skyreels.ai/ ; Mureka – https://www.mureka.ai/