Summary of Key Points
After returning from his hiatus, DingTalk's founder, Wu Zhao, launched the AI-powered office product ONE. The goal was to use AI to organize scattered work information (from group chats, to-do lists, meetings, etc.) and create an efficient experience where "tasks find users." However, ONE faced multiple conflicting objectives: reducing user burden, transforming the product, boosting team morale, and generating commercial revenue. Additionally, there were internal contradictions regarding its positioning (pleasing bosses versus employees) and the high-pressure iteration process that led to a lack of foundational development. Within just 10 months, ONE evolved from a promising AI innovation to a transitional feature before being ultimately discontinued. Its failure highlights the common challenges faced by tech companies during their AI transformation efforts—overambition and a rush for quick success.
ONE Project: The Dream of "Tasks Finding Users" in the AI Era
The concept behind ONE was straightforward: previously, users had to manually search through group chats for tasks, check meeting notes, and follow up on approval processes. Now, AI would sort this information into prioritized cards and push them to users, shifting the focus from "users finding tasks" to "tasks finding users." For example, after a meeting, ONE would automatically send out meeting minutes, tasks, and relevant documents. Wu Zhao hoped that ONE would become the new "home page" for DingTalk's AI capabilities, demonstrating that DingTalk was not just an old-fashioned office software but could keep up with the AI trend.
The Four Major Objectives
From the outset, ONE was designed to fulfill four primary goals:
1. User Focus: To reduce the workload on employees by organizing scattered work information.
2. Product Development: To provide DingTalk with a new entry point into the AI era, rather than merely integrating AI within existing features.
3. Organizational Goals: As Wu Zhao returned to lead the company, ONE was intended to be a success that would strengthen the team and change how the outside world viewed DingTalk.
4. Commercial Success: Since AI implementation required costs (such as model licensing), it needed to be integrated into DingTalk's services to demonstrate its value and generate revenue for the company.
These four objectives pulled ONE in different directions. For instance, achieving quick organizational results might have come at the expense of user experience, and inserting advertisements into work processes could have annoyed users.
Inherent Contradictions
The very positioning of ONE was inherently contradictory:
- Pleasing Both Bosses and Employees: DingTalk has two types of users: bosses who need to oversee employee productivity and transparency, and employees who want freedom from constant monitoring. ONE aimed to serve both groups but failed to satisfy either.
- DingTalk's Core Values vs. AI Vision: DingTalk is designed to assist bosses in managing employees (e.g., with features like "unread notifications" and "DING" reminders), while ONE claimed to reduce the burden on employees by filtering out unnecessary messages. However, these features caused backlash, as employees felt their privacy was being侵犯.
- An Impossible Balance: The product had to cater to a wide range of users, encourage frequent usage, and generate revenue. It also had to be user-friendly (e.g., avoiding distracting ads in work-related modules).
High-Pressure Iteration
The ONE team adopted a "daily delivery" approach, where requirements from bosses had to be met by the end of the day. This pressure prevented them from investing time in essential foundational work, such as personalized recommendations and feedback loops. As a result, the product had many features on the surface but lacked stability: inaccurate AI recommendations and user dissatisfaction. Employees worked long hours with high turnover (only three people stayed for more than three months), and one employee even fainted from exhaustion.
Lessons from ONE's Failure
ONE's failure is not a sign of the team's lack of effort but rather of focusing on the wrong priorities. It highlights three important lessons for the industry:
1. Don't Overload Products with Too Many Goals: Trying to address multiple issues (user needs, organizational goals, and revenue generation) simultaneously is like trying to drive a car in four different directions at once—it will inevitably lead to failure.
2. Balance Traditional Values with New Vision: When transforming an old product with AI, don't overlook its core strengths. Forcing changes that conflict with the product's nature (e.g., trying to help employees defy management) can make users feel disengaged.
3. Take Time to Build a Solid Foundation: AI products require time to refine user experiences and train models. Rushing to achieve results often results in superficial success.
In summary, the fate of ONE serves as a reminder for tech companies: in the AI revolution, don't get caught up in hype and short-term KPIs; focus on understanding what users truly need.