Kun Jing, Genspark: Dancing with Microsoft Agent 365, Making AI Agents Everyone's Work Partner

Remember "Vibe Working," the concept Genspark CEO Kun Jing introduced in his earlier columns? That state of working with AI not as an adversary but as a默契 partner — it's now landing through a more powerful ecosystem.

At this week's Microsoft Ignite conference, Genspark officially announced its entry into the Microsoft Agent 365 ecosystem. This isn't a simple API integration. It's a significant signal in the evolution of AI agents: AI is moving from standalone "chat windows" into the core productivity platforms used by hundreds of millions of knowledge workers worldwide.

Genspark, live for just one year, has already surpassed $50 million in annual recurring revenue (ARR) while maintaining exceptionally high user retention. The driving force behind this is Jing's insistence on a philosophy of AI interaction — don't just chat with robots; talk directly to delivered results.

BlueRun Ventures is Genspark's angel and Series A investor. We share the same wavelength as Jing. The future is here, and it's becoming tangible. Below is an excerpt from Jing's interview about this partnership, compiled by BlueRun. Enjoy reading:

---

"We're pushing for a shift in perspective: evolving AI from a mere 'chat counterpart' into a 'daily work partner' that solves problems, and actually getting things done on a unified platform." This has been the vision of Kun Jing, Genspark's co-founder, and the platform carrying this vision is Genspark Super Agent. The AI startup he founded with Kay Zhu in 2024 now has a presence extending to Singapore and Japan. "We want users to be able to talk directly to 'delivered outcomes,' rather than chatting at a robot. It's a fundamentally different user experience," Jing says, contrasting Genspark with the LLMs commonly used today for writing or coding. On Genspark, the interaction logic is intuitive: "Your conversation partner is the deliverable itself." If it's a slide deck, you talk directly to the slides; if it's a spreadsheet, you communicate directly with the table. This "results-oriented" interaction has allowed Genspark to rapidly accumulate more than 80 native agents, handling everything from video editing and full website builds to complex marketing collateral design and even restaurant reservations — all in one place. And now, this "direct delivery" capability is being packaged into Microsoft's ecosystem. At Ignite, Genspark announced it would join Microsoft Agent 365 alongside top-tier players like Cognition and Glean. What does this mean? It means Genspark's Super Agents will no longer be isolated islands. Jing explains: "We want Genspark to be deeply embedded in the Microsoft ecosystem. Knowledge workers of the future won't need to leave their familiar Microsoft 365 interface to tap into Genspark's problem-solving capabilities."

According to Microsoft's 2025 Work Trend Index, enterprises worldwide are deploying thousands of agents to automate routine tasks. IDC projects this number will surge to 1.3 billion by 2028. But a challenge comes with it: how do you manage these "digital employees"? Who created them, what data can they access, are their actions secure — these questions become critical. Unmanaged agents can become invisible threats, leading to data leaks or compliance violations. Agent 365 is the "command tower" built for this purpose. It lets enterprises manage agents the way they manage employees — granting them secure permissions, monitoring their operational compliance. This addresses the biggest concern holding businesses back from embracing AI. What does this mean for real business scenarios? Imagine a busy marketing director who urgently needs to prepare a multimedia proposal for the board. In the past, this might have meant days of grinding. Now, they can go straight into Microsoft Agent 365, select Genspark Super Agent, and add it to their team. Because this agent comes with a built-in Microsoft Entra Agent ID, it can securely connect to internal company resources. The director can instruct the agent to launch research and generate presentations, write documents, build websites, produce videos and posters, and more. The agent can then adjust and update these materials based on email feedback or discussions in Teams meetings. Throughout the entire process, Agent 365 silently logs and reviews everything in the background, ensuring full compliance with company policies. Jing is convinced this combination will create genuine "magic moments" for users. Take PPT creation, for example. Genspark's AI Slides feature was designed from the ground up with Microsoft PowerPoint compatibility in mind. "We went through four iterations on the export function," Jing says. "For those already accustomed to PowerPoint formats or who need to account for company workflows, their exported files can seamlessly adapt to PowerPoint." Still, Jing offers a sobering dose of cold water amid the frenzied tech hype. No matter how powerful AI becomes, "people" remain the indispensable core of any workflow. "In our philosophy, AI can beautifully handle 80% to 90% of the foundational work. But that remaining 5% to 10% — what we call 'the last mile' — must be completed by humans," Jing emphasizes. "You need to verify, to gatekeep, to inject soul."

For Genspark — founded just a year ago, with its product launched only in April 2025 — the Microsoft partnership is a leap forward. Despite a team of just 30 people, Genspark's goal is sharply focused: help knowledge workers achieve more. Jing notes that Microsoft provides a vast channel for reaching enterprise users, a segment startups often struggle to cover independently. "It's a massive marketplace bringing together all kinds of excellent software," he adds. With this clear understanding of where human-machine collaboration boundaries lie, and an extreme focus on solving knowledge worker pain points at the lowest possible cost, Genspark has in just one year joined the ranks of the fastest-growing AI startups globally, reaching $50 million in annual recurring revenue (ARR). Genspark's agents are powered by a hybrid architecture of eight large models — including well-known models like OpenAI's GPT and Anthropic's Claude — and with Microsoft Azure's computational backing, Genspark can dynamically call the most suitable model for each task, ensuring users get optimal results across different workflows. By the core survival metrics for AI startups, Genspark is unquestionably in the fast lane. Beyond the ARR figure mentioned earlier, Genspark's paid retention rate in its first month of launch reached 88% to 92%, confirming exceptionally high user loyalty. At the same time, the company's capital efficiency is remarkable: average monthly burn through Q2 2025 (ending June 30) was kept below $1 million. Jing reveals that Japan, the United States, and South Korea are currently Genspark's largest paid subscription markets. In his view, the industry's ceiling remains far from touched. He believes the AI revolution has so far penetrated "less than 1% of society," and the quality of work AI can deliver still has enormous room for improvement. "Two years from now, I foresee a fundamental shift in the AI market's center of gravity: from serving the developer community to truly serving the broad population of knowledge workers," Jing concludes.

In this era of AI shifting from talk to action, Genspark has already sprinted to the front of the pack.

Genspark CEO Kun Jing: I Saw AGI, and I Hope I'm Wrong | BlueRun Family Headlines Kun Jing's Latest Column: Multi-Model and Multi-Agent Is the Future Genspark CEO Kun Jing: Don't Want to Be Replaced by AI? Here's What I Suggest You Start Doing Now