Merge CEO Shensi Ding: AI Fluency Starts at the Top

In today's world where AI tools are increasingly widespread, many companies are still exploring how to truly integrate AI into daily workflows. Merge CEO Shensi Ding, in a recent sharing, revealed a key trend through her personal experience: AI fluency must start at the top. She first used Codex after Astra's launch, and this tool completely changed her understanding of AI capabilities.

Astra: From "Executing Tasks" to "Filling in the Blanks"

Ding mentioned that Astra not only executes tasks one by one, but does so exceptionally well, and uses common sense to fill in blanks she didn't explicitly specify. This capability breaks through the limitations of traditional automation tools—it no longer merely responds passively to instructions, but actively understands intent. For team members without engineering backgrounds, this means they can independently build the tools they need without waiting for technical colleagues' assistance.

Codex is a code generation model launched by OpenAI, initially provided as an API and later integrated into tools like GitHub Copilot. Based on a code fine-tuned version of GPT-3, it can generate code snippets in multiple programming languages from natural language descriptions. Codex's emergence marked AI's leap from "understanding language" to "generating executable content," laying the foundation for subsequent more intelligent automation tools like Astra. Ding's first experience with Codex was precisely the starting point of her cognitive shift—she discovered that AI could not only answer questions but also directly produce usable code, which led her to begin reevaluating AI's role in workflows.

Productivity Liberation for Non-Engineering Teams

Ding particularly emphasized that Astra is enormously helpful to non-engineering team members. In the past, they had to queue and wait for engineering resources to implement a workflow improvement idea; now, they can immediately create solutions for themselves. This "self-sufficient" model significantly reduces friction in cross-departmental collaboration and allows innovative ideas to be quickly implemented.

How Top-Level Demonstration Drives AI Adoption

As CEO, Ding's personal practice sends a clear signal: AI fluency is not an elective for frontline employees, but a required course for leadership. When leaders are the first to use AI tools and share success stories, they drive the entire organization's trust in and willingness to try AI. This top-down demonstration effect is more effective than any mandatory training.

AI adoption refers to the process by which organizations integrate artificial intelligence technology into business processes, typically involving technology deployment, employee training, and cultural change. Research shows that support from senior management is one of the key factors for successful AI adoption. Through personal demonstration, leaders can reduce employees' resistance to unfamiliar technology and convey the signal that "AI is a tool, not a threat." Ding's practice as CEO is a typical case of this "top-down" adoption model: her usage behavior itself becomes the most powerful advocacy within the organization, more capable of inspiring employees' willingness to try than mere administrative orders or training courses.

Implications for Enterprises

Ding's experience points to a broader trend: AI tools are shifting from "expert-only" to "available to everyone." If companies want to maximize AI investment returns, they should prioritize cultivating AI literacy among management and choose tools that can understand context and possess common-sense reasoning capabilities. Merge's practice shows that when non-technical members can also independently build workflows, the organization's innovation speed and responsiveness will gain substantial improvement.

One-line Summary

Merge CEO Shensi Ding shares: AI fluency requires top-level demonstration, Astra helps non-engineering teams independently build workflows.

Paragraph Summary

Merge CEO Shensi Ding (Shensi Ding), through personal experience, points out that AI fluency must start with top-level leaders. After her first use of Codex, her understanding of AI capabilities underwent a shift, and the Astra tool not only executes tasks but also uses common sense to fill in unspecified blanks, breaking through traditional automation limitations. This is significant for non-engineering teams: members no longer need to wait for engineering resources and can independently build workflows, reducing collaboration friction and accelerating innovation implementation. As CEO, Ding's demonstration conveys the signal that AI is a required course for leadership, and the top-down driving effect is more effective than mandatory training. Enterprises should prioritize cultivating management's AI literacy and choose tools with contextual understanding and common-sense reasoning to maximize AI investment returns.