Author: Wang Ruihao | Editor: Li Yuchen
An executive at a traditional automaker once recalled to Leiphone that during the pandemic, the head of his company visited Bosch in Suzhou.
At that time, automakers were busy securing supply. According to his memory, this CEO-level executive waited there for several days regarding the supply of an active safety system, but ultimately failed to meet with the person in charge and did not secure the supply.
Upon leaving Bosch, he sighed deeply: "Who is actually Party A, and who is Party B?"
He remembered this incident, and it prompted this traditional automaker to make a firm decision: the most critical components must be developed in-house.
The story of traditional automakers pursuing autonomous driving often begins with such determination. However, as they progressed, some companies achieved mass production, while others only faced one organizational restructuring after another.
Money was spent, and talent was hired. Why still no results? Several autonomous driving heads and experts who have long served automakers pointed to two key issues: whether funding can be sustained, and whether those who truly understand the technology are empowered to lead.
01Hesitation in Investment, Suspicion in Personnel Management
The automotive industry is not short of grand statements about self-developed autonomous driving. Once truly initiated, management at large enterprises easily reverts to familiar patterns: invest a little first, see the effect, then decide on the next tranche of funding.
Wang He (pseudonym), a headhunter who has deeply cultivated the automotive sector for years, observed this shift through the experiences of technical experts he has served. In his observation, initial investments lacked a clear overall budget plan, adopting a strategy of "invest a little to try, then gradually add more."
When results were slow to appear, management began questioning where the money was going. Subsequently, original R&D issues transformed into approval bottlenecks.
"Once monitoring starts, all approvals get stuck; once stuck, the entire pace slows down, eventually leading to step-by-step delays," said Wang He.
In his view, this management style trapped autonomous driving R&D in a vicious cycle: poor results led to tightened investment; tightened investment slowed progress further; slower progress made it harder to secure the next round of support.
For automakers accustomed to calculating costs on a project-by-project basis, this approach seems reasonable. However, autonomous driving requires upfront heavy investment to build system capabilities, followed by continuous iteration.
Chen Xiaotong (pseudonym), an autonomous driving executive who previously served a traditional automaker, noted that traditional OEMs often lack AI capabilities and infrastructure for closed-loop data systems. After recruiting algorithm talent, data collection, training, testing, and engineering delivery must also keep up. If any link is missing, single-point breakthroughs struggle to translate into whole-vehicle experiences.
Being willing to pay high salaries for only a few experts but unwilling to fund an entire team is another form of hesitation.
"They can afford high wages for three or five people but hesitate to do so for fifty," Chen Xiaotong said. "This task inherently requires 1,000 people, or at least 500. You definitely cannot drive it forward with just five people."
The scale and pace of investment must match the goals. Without this understanding, management easily compresses foundational investments while demanding teams quickly catch up with industry leaders.
More troublesome is that some companies indeed spent money and recruited many excellent talents, yet still failed to deliver products.
The problem lies in the middle layer.
Wang He has encountered several technical geniuses who left automakers. In his description, these individuals came from top universities like Tsinghua and Peking University, received competitive compensation, but were managed by "senior personnel" who rose through the old system based on tenure. Some of these seniors had educational backgrounds and technical training significantly distant from the knowledge structures required for new business areas.
Degrees certainly do not directly equate to capability. But as technology shifts toward AI, if those controlling resources and deciding directions fail to update their cognition, even a team full of PhDs will struggle to automatically grow an autonomous driving system.
A common issue among these "seniors" is their inability to utilize talent effectively.
Externally recruited technical experts often excel in specific technical directions and need to be integrated into a clear overall architecture. Upper management must identify current bottlenecks, determine who should break through them, and define how breakthroughs enter mass production systems.
If each department develops its own separate solutions and technical routes fluctuate repeatedly, research outcomes may never find a landing point. When performance reviews arrive, managers inversely question: What exactly have you done every day?
People are busy, producing numerous papers and reports, yet limited results actually reach the vehicle. Companies pay high salaries, employees consume time, and both sides feel shortchanged.
There is an even deeper issue: some "seniors" are unwilling to let "newcomers" succeed.
In Wang He's view, certain "seniors" fear that newcomers' stronger technical abilities and resumes, combined with tangible achievements, would challenge their positions.
He described the organizational status of one company as follows: Newcomers are responsible for business outcomes, while recruitment, personnel management, and administrative control remain in the hands of seniors. "The task is handed to you, but resources may not follow."
Consequently, technical debates easily evolve into power struggles. Whose solution is adopted, whose team receives budget, and who gains greater voice. Before results emerge, internal battles over ownership of potential achievements begin.
Chen Xiaotong's judgment on this is direct: "I believe one of the biggest problems with traditional automakers is the excessively slow turnover of talent."
He believes there is nothing wrong with existing employees continuing to enjoy company benefits, but key technical positions must be taken over by those truly capable.
This also explains why hiring a star executive and recruiting batches of graduates from prestigious schools often proves ineffective.
Chen Xiaotong emphasizes the importance of tiers: leaders guiding the way in front, intermediaries in the middle who understand direction and decompose problems, driving executors below. Missing this middle layer makes it difficult for heads to implement direction, and young talents' capabilities fail to connect to products.
Some automakers also prefer having multiple internal teams compete against each other and suppliers, hoping to pick a winner. Chen Xiaotong believes this leads to extremely high system complexity and prevents R&D teams from focusing.
Ultimately, energy originally intended for model training and solving mass production issues is wasted on proving oneself, competing for budgets, and coordinating departments.
Money was not fully invested, and people were not truly utilized. Autonomous driving thus remained trapped within the organization.
02What Did Those Who Succeeded Do Right?
Traditional automakers have not been without success. SAIC Motor is a prime example.
Around 2021, SAIC introduced a technical executive from Huawei's Auto BU, appointing him as Chief Scientist of the R&D General Institute to lead the autonomous driving development for Rising Auto.
Leiphone learned that the team led by this executive had produced results at SAIC, but later faced marginalization, and the team could not continue. "It is quite regrettable," said an insider.
According to public reports, SAIC Rising's PP-CEM team implemented the Rising Pilot solution in the R7 and F7 models, subsequently facing role transfers and adjustments. SAIC's early self-development was indeed just one step away from mass production.
This case invites scrutiny: Why couldn't a team that had already delivered results continue building on their accumulation? Technical capabilities formed can still be interrupted by organizational changes.
ZEEKR provides another sample.
It is the enterprise within the traditional automaker system that truly succeeded in self-developed autonomous driving.
In late 2021, Chen Qi, founder of Huawei's autonomous driving team and former director of autonomous driving R&D, joined ZEEKR. By the end of 2023, the ZEEKR 007 launched with self-developed Haohan Intelligent Driving; in 2024, capabilities continued expanding from highway to urban scenarios.
The successful mass production of ZEEKR's self-developed autonomous driving credited primarily to An Conghui, then CEO of ZEEKR, providing unconditional trust to his subordinate teams.
A Geely executive once described An Conghui's management style: "Director An holds an umbrella above you, letting you work freely."
This trust is particularly critical. Internal relationships in traditional automakers are complex; new teams may face suspicion before delivering results. Various doubts reaching upper management can easily shake judgments, causing teams to lose support for continuation before winning their first battle.
An Conghui was willing to filter out noise and strongly support Chen Qi's team because he believed in Chen Qi's ability and understood how difficult it is to master autonomous driving.
It is understood that the team built under Chen Qi's leadership was already Geely's third generation of self-developed autonomous driving teams, with An Conghui initiating each generation. Having experienced two previous setbacks, he clearly understood the difficulty of self-development and better appreciated the patience and support needed for a new team.
(For insights into the evolution of Geely's three generations of autonomous driving self-development, welcome to add the author's WeChat jinyuan-0428 for discussion)
Luckily, the third time proved lucky, allowing An Conghui's obsession with autonomous driving self-development to resonate through Chen Qi's generation team.
In retrospect, ZEEKR's successful transition to mass production relied on two individuals: An Conghui provided trust and authorization; Chen Qi organized talent, data, tools, and development processes.
Recently, Changan Automobile successfully mass-produced the "Tianshu Navigation" autonomous driving system after three years, with the core factor again being people.
Changan's autonomous driving head is Tao Ji, a veteran of Baidu's autonomous driving division and a genuine technical expert.
Changan's success in autonomous driving similarly stems from the combination of determination, authorization, and the leader's capability.
In September 2023, Tao Ji chose to join Changan Automobile. Prior to this, he hesitated between starting a venture and joining an OEM.
It is reported that Wang Jun, then President of Changan Automobile, adopted a very humble posture when recruiting Tao Ji. He told Tao Ji: "Come help me today; tomorrow I will definitely support your implementation of embodied intelligence at Changan Automobile."
This statement moved Tao Ji. Subsequently, Wang Jun began fulfilling his promise, repeatedly mentioning embodied intelligence at strategic meetings of various scales, pushing the initiative into Changan's plans.
After entering actual work, Tao Ji gained significant autonomy in recruitment, team building, and determining technical directions. Behind this was the full authorization and trust from Changan's then Chairman and President.
Viewing the cases of ZEEKR and Changan together, the determination of the top leader and the capability of the technical head are indispensable. The former must be willing to invest, dare to use talent, and bear the risk of trial and error; the latter must deliver products and live up to this trust.
The greatest crux for traditional automakers in autonomous driving lies in whether they can empower those who truly understand the technology to lead, giving them sufficient time to achieve success.