Ctrip promotes product-R&D integration; ByteDance, Alibaba Cloud and others experiment with AI-native organizations
According to a self-media report, Ctrip held an all-hands meeting last week and announced the promotion of "product-R&D integration," regrouping product, operations, front-end and back-end, algorithms and testing by project, with a single DRI managing them uniformly, and reporting and performance evaluation possibly breaking functional boundaries. The article also mentions that Toutiao uniformly configures R&D for three platforms by business module, that Lenovo China's software team has popularized "full-stack engineers + 5 agents," and that Alibaba Cloud's security product mixed-team pilot shortened the average requirement processing cycle from 50.9 days to 13.8 days. The author believes that non-technical leaders find it difficult to evaluate code quality and architecture, and that if performance, promotion and professional development mechanisms are not changed, tearing down departmental walls may turn into building project walls.
Why read this
AI score 52
The article compares the organizational adjustments at Ctrip, ByteDance, Lenovo and Alibaba Cloud, presenting the resistance at the level of processes and performance mechanisms when AI is implemented.
Topics
Body · AI translation
[[**According to reports, Ctrip is promoting "product-R&D integration," reorganizing product, operations, and R&D by project under DRI management; ByteDance, Lenovo, and Alibaba Cloud are also experimenting with AI-native organizations. AI is forcing big tech companies to tear down departmental walls, but without changes to performance and management mechanisms, implementation will be difficult.**
**Key Points**
**1. Ctrip reorganizes by project system, with DRIs in unified charge**
According to reports, product, operations, front-end and back-end, algorithm, and testing staff are teamed up by project, with about 1-2 people per job type; the person in charge can come from product, operations, or technology, and reporting and performance evaluations may break functional boundaries.
**2. ByteDance, Lenovo, and Alibaba Cloud have tested AI-native organizations**
Today's Headlines configures R&D for all three ends by business module; Lenovo popularizes "full-stack engineers + 5 agents"; after Alibaba Cloud's security products were mixed-staffed, the pilot requirement cycle dropped from 50.9 days to 13.8 days.
**3. AI lets one person cover multiple roles, while old processes slow delivery**
**4. Performance, promotion, and management authority determine success or failure**
Non-technical leaders find it hard to evaluate code quality and architecture; if professional training, basic technology maintenance, and the performance system are not changed, tearing down departmental walls may turn into building project walls, with middle management facing the greatest resistance. 2026-11-14 09:59 ByteDance, Alibaba, Ctrip begin tearing down departmental walls Everyone Is a Product Manager Quick Overview
This article comes from the WeChat public account: Everyone Is a Product Manager, author: Guai Ge
Product managers give programmers performance reviews, and programmers report to operations.
This kind of scene, which sounds a bit surreal at big tech companies, may really be about to happen.
According to a self-media report, Ctrip held an all-hands meeting last week and announced the start of promoting an organizational reform called "product-R&D integration."
How exactly will it change? People originally from different departments—product, operations, front-end, back-end, algorithm, testing, and so on—will all be reorganized into teams by project. In a team, each job type may have only 1 to 2 people, managed uniformly by one direct responsible individual (DRI).
Even reporting relationships and performance assessments may break the original functional boundaries.
Ctrip's approach is quite radical: everyone is full-stack, everyone is a product manager.
To be honest, Ctrip's step is a bit big...
Think about the past: from a product requirement being proposed to going live, it had to go through a series of steps including operations proposing the requirement, product reviewing the requirement, writing the PRD and drawing prototypes, design, technical evaluation, R&D scheduling, and testing acceptance.
When encountering complex requirements, several departments also had to be coordinated. Everyone held the requirement meeting, review meeting, and scheduling meeting, and then another progress sync meeting.
Each department had its own KPIs and priorities: product wanted to launch as soon as possible, technology worried about architecture and stability, and operations were eager to launch, validate, and get results. When problems arose, it was easy to blame each other.
Sometimes a feature itself might need only two days of development, but communication and coordination alone consumed one or two weeks.
Ctrip's reform this time is clearly intended to solve these problems.
But what really interests Guai Ge is that this reform may break traditional organizational relationships and collaboration together.
Can a product manager manage engineers? Does an operations leader have the ability to evaluate the quality of technical staff's work?
Big tech companies that used to divide teams by function will probably have to re-study how to do performance assessments.
Moreover, there have indeed been quite a few big tech companies taking similar actions recently.
At the end of October this year, according to media reports, ByteDance's Today's Headlines adjusted its R&D organization, changing the front-end, client, and server teams originally divided by function into unified configuration by business module.
One business module is managed by the same leader for R&D across all three ends, directly responsible for technical delivery results.
This is equivalent to reorganizing R&D staff originally scattered across different departments into teams capable of independent delivery.
Lenovo has gone further.
In September this year, the "AI-Native Organization Transformation Guide" jointly released by Lenovo and Renmin University of China disclosed that full-stack engineers had already become widespread across all software teams in Lenovo China.
Work that previously required multiple roles such as front-end, back-end, testing, and operations is now being attempted by Lenovo through a "full-stack engineer + 5 agents" approach.
If Lenovo is mainly changing the division of roles, then Alibaba Cloud has already begun studying how the entire R&D organization should change.
On November 11, the Alibaba Cloud developer community published the CTO product-R&D line's AI Native practice over the past three months.
Alibaba Cloud explicitly mentioned allowing teams, during the experimental stage, to cross original role boundaries, explore mixed staffing of roles such as product, R&D, and security, and new mechanisms for humans and Agents to collaborate.
In one security product project, Alibaba Cloud integrated product, R&D, and security personnel into the same team, with each requirement handled by one Owner from goal definition through result acceptance.
According to pilot data disclosed by Alibaba Cloud, the average requirement processing cycle was shortened from 50.9 days to 13.8 days.
This data cannot yet represent all of Alibaba, but at least it is somewhat persuasive.
Actually, many big tech companies have long launched multiple rounds of internal organizational reform.
The reason is not complicated.
In the past, internet companies emphasized professional division of labor because every role had a high technical threshold.
Product managers were responsible for figuring out what to do, designers for aesthetics and experience, programmers for writing code, testing for finding bugs, and operations for bringing products to market.
To manage so many people, different functional departments, leaders at different levels, and a large number of processes were needed to coordinate work.
This organizational model has become standard for domestic internet companies and has been running for more than a decade.
But as technologies like AI Coding and Agents become more mature, many things are starting to change.
Here comes the question.
If one person can do the work of several roles, do companies still need to keep so many functional boundaries, approval steps, and management layers?
I think this question will probably put pressure on quite a few managers at big companies.
That's a bit awkward.
Of course, departmental walls can't just be torn down at will.
For example, if you put an operations lead in charge of an R&D team, he may understand business growth very well, but he may not be able to accurately judge code quality, technical architecture, and system stability.
If you let a product manager develop features with AI on their own, building a demo may be easy, but actually launching a system involving payments, orders, and user privacy is not so simple.
What's even more troublesome is performance evaluation and promotion.
If employees are no longer divided according to traditional functions, who evaluates professional ability? Who is responsible for training newcomers? And who maintains the basic technical capabilities needed by multiple project teams over the long term?
If these problems are not handled well, the original departmental walls may be torn down, only for project walls to be built one after another in the end.
There may even be an awkward result: departments are merged, but the number of managers does not decrease, and there are even more meetings.
So I think whether Ctrip's reform this time can succeed ultimately depends on whether the management mechanism can change along with it.
If it merely pulls product, operations, and R&D into the same group chat and swaps in a new person in charge, it probably won't solve much.
For product managers, this round of change is also worth being vigilant about.
In particular, those product managers mainly responsible for writing requirement documents, passing requirements along, and coordinating schedules may see their future career space become narrower and narrower.
But people who truly understand users, are familiar with the business, can make product judgments, and are willing to take responsibility for the final result may instead be more valuable.
Likewise, engineers also need to understand the business better, and operations staff need stronger product thinking. Everyone has to learn to break through the original boundaries of their roles.
In the past, a team of a few dozen people might need several functional leaders. Each department had its own reporting relationships, approval authority, and performance system.
This is probably harder to push forward than getting employees to learn AI Coding.
It's unclear what the final implementation effect of Ctrip's reform this time will be, but the direction is certainly right.
Who knows, one day the major internet companies may start copying it at scale.
Has published 239 articles on Huxiu. This content is published with the author's authorization. The views represent only the author and do not represent Huxiu's position. If you have objections or complaints about this manuscript, please contact tougao@huxiu.com. Everyone who is changing and wants to change the world is on the Huxiu APP[ ]
AI Trends