Rebuilding Growth with AI: The Practice of New Channel Education Group | 2026 AI Partner · Beijing Yizhuang AI+ Industry Conference

Background

Xintong Education is not a startup built from scratch around AI, but a 30-year-old education group with more than 30 subsidiaries and over 2,000 employees. Its legacy business serves the international development needs of China’s middle- and high-income families through study planning, employment planning, and international curriculum services. That context matters because the speech is about a harder problem than simple product innovation: how a large traditional organization absorbs AI without abandoning its accumulated assets, teams, and operating logic.

Ma Yawei frames the challenge in contrast to entrepreneurs who can shut down one model and restart around AGI. Her company could not do that. It had to pursue simultaneous optimization of old business and exploration of new growth, which is what makes the case especially relevant to incumbent firms.

The Four Stages of Transformation

Ma presents the transformation as a staged progression rather than a one-time upgrade:

1. 2021–2022: Digital infrastructure

   • Xintong spent two years consolidating data and systems.
   • She compares this to turning fragmented “provincial roads” into a highway for AGI.
   • The implication is that AI capability depends on prior digital order; without it, later experimentation lacks operational grounding.

2. 2023–2024: AI Must

   • This phase focused on making AI mandatory in principle, but not through pure top-down enforcement.
   • The company brought in external advisors, ran company-wide open classes, encouraged employees who had already used AI to teach peers, built benchmark pilot groups, and tested product prototypes at Harvard and Stanford.
   • The deeper move here was changing AI from a management order into a shared belief.

3. 2025: AI in All

   • AI became embedded in all key processes.
   • Employees received AI partners, digital employees, intelligent tools, and intelligent platforms.
   • Success is described not just as tooling adoption but as a cognitive shift: employees facing a problem now instinctively ask how AI can handle it.

4. 2026: AI Native

   • The next phase is not adding AI into existing workflows but designing new businesses with AI-native logic from the start.
   • This marks the shift from tool empowerment to organizational genetic reconstruction.

Key Points

1. The real bottleneck is organizational transmission

The speech’s strongest claim is that the hardest part of AI transformation is not deciding strategy but carrying it from the boardroom to the frontline. Ma explicitly says founders often have little trouble recognizing future trends; the difficult part is ensuring that strategy becomes lived operational behavior.

Her answer is notable because it rejects simple managerial coercion. She argues that:

• commands do not solve employee indifference,
• preaching does not create participation,
• even basic incentives are insufficient.

Instead, Xintong used carefully designed internal social proof. External experts opened the conversation, but frontline employees became the main evangelists. That choice lowered the perceived barrier to entry and made AI feel accessible rather than elite. The insight is powerful: consensus is built horizontally, not just imposed vertically.

2. Pilot design matters more than abstract enthusiasm

Xintong’s method for choosing AI use cases is highly pragmatic. It prioritized:

• scenarios with the largest number of employees,
• scenarios with the longest time consumption,
• scenarios that were internal rather than directly customer-facing.

This is a sophisticated sequencing choice. By starting with internal efficiency cases, the company reduced external risk while maximizing visible employee benefit. It made AI transformation feel immediately useful rather than strategically distant. The article suggests that adoption accelerates when employees can directly experience saved time and lower repetitive workload.

3. Data accumulation becomes strategic advantage under AI

The Harvard and Stanford product validation phase gave Xintong more than confidence; it clarified where its advantage lay. Ma says the company realized that its decades of language learning data gave it unique strength in vertical AI products. Later, she mentions:

• 500,000+ study-abroad users,
• hundreds of thousands of global intelligent cases,
• accumulated trajectories including admission, rejection, and employment outcomes.

This means Xintong’s transformation is not based on generic AI use alone. It is based on turning historical industry data into proprietary intelligence systems. In her analogy, this is like giving a doctor access to an enormous case center. The article implies that traditional firms are often richer in domain data than they realize; AI unlocks the value of that inheritance.

4. AI is being used to formalize tacit service knowledge

One of the most concrete examples is the creation of holographic user profiles. In education consulting and teaching services, personalized needs are difficult to aggregate and record. AI labeling and interaction tools convert fragmented customer conversations from phone, WeChat, and other channels into structured, evolving profiles.

This is significant because it addresses a classic service-business problem: high-value work is often trapped in employee memory rather than institutional systems. Xintong’s AI tools turn interaction residue into reusable organizational knowledge, improving service continuity and personalization.

Results and Evidence

The speech provides several indicators of progress:

• revenue and labor efficiency increased even without headcount growth, and in some cases with fewer people;
• a second growth curve reached meaningful revenue scale in 2025;
• Xintong won the 2025 Harvard Business Review annual excellence award for digital transformation;
• three products were selected for display at the 2026 World Digital Education Conference;
• the AI-native IELTS product reached 300,000 learners and 20 million RMB in revenue by the end of 2025;
• token usage rose from 300 million in 2024 to 12 billion the following year, signaling deep internal adoption.

These details matter because they show that the company is not describing a vision-only program. It links organizational change to measurable product and operating outcomes.

Significance

The article’s broader importance lies in its model for incumbents. Xintong does not present AI transformation as a sudden reinvention but as a sequenced organizational learning process:

• build infrastructure,
• create common language,
• prove value in internal workflows,
• scale into all processes,
• then launch AI-native businesses.

The key insight is that AI transformation is cultural and architectural before it is commercial. New businesses such as the AI-managed IELTS and Chinese-learning products emerged only after the company had already changed how people worked and how knowledge was stored.

Ma’s final management lessons reinforce this:

• combine external expertise with internal talent activation,
• use attraction rather than command,
• maintain strategic alignment while allowing some waste and experimentation,
• create internal incubation funding for AI-plus-industry exploration.

In short, the article argues that traditional firms already possess the raw materials for AI-era renewal—industry knowledge, data, customer relationships, and organizational scale—but they must first solve the problem of turning executive strategy into everyday behavior.