An AI-native org isn't everyone using AI — it's assuming the org doesn't exist and rebuilding it
Lockin merged HR and IT into HARO, put a technical founder-type in charge of people and AI, gave circle leads the power to split the bonus pool, and now requires proof that AI can't do the job before anyone gets hired.
The video won't play here. Listen to the audio instead:
The argument · tap a timestamp to hear it
AI-native starts by assuming the organization doesn't exist
Chen Bin's definition of an AI-native organization is not ‘everyone should use AI well’ but starting from AI and asking again: can AI do this thing, and as far as possible don't have a person do it. The harsher step is to assume the original organization and processes don't exist — ‘they shouldn't even exist’ — and go back to the state of a one-person startup to redesign roles and processes. So this is not accelerating existing processes; it's tearing the processes down first and rebuilding. He concedes this is mainly aimed at the core org builders, not a slogan shouted at everyone.
— Chen BinThe HARO lead comes from tech, not HR
Lockin merged HR and IT into HARO (Human AI Resource Office), but Chen Bin stresses it is not an ordinary HR+IT combo: the lead must understand the business, and the person chosen is an old partner who co-founded with him and has done technology, product, CEO, R&D, brand and IT. The logic is inverted — the old HR lead leaned toward people and was insensitive to AI, so now a tech-oriented person leads, with HR BPs who are sensitive to people assigned to him. Under this department sit the AIO (the former IT team, which will do Agent Builder work) on a solid line, and FDEs reporting on a dotted line.
— Chen BinAgent Builder and FDE are two different people in the short term
Chen Bin draws the line clearly: an Agent Builder leans technical and can build an Agent; an FDE has to go into the business front line and map out supply, planning and procurement processes, and is best a business expert. Long term one person might cover it; short term it must be two kinds of people working together. They did not copy Anker's approach of pulling people out of work and locking them in a room for a month, because they don't have that much slack; instead people from the business spend 30% to 50% of their energy doing AI-ification alongside their day jobs. The company also set up an AI proficiency ladder from L1 to L4, with L2 and above called ‘AI 起航员’ (AI navigators) and above that ‘领航员’ (pilots).
— Chen BinCircles erase functional power; circle leads get the bonus pool
Lockin's circles differ from project teams and Huawei's matrix: a matrix is product line plus function on two axes, while a circle is a small team around an end-to-end value loop, deliberately erasing functional power and boundaries so the circle lead's power grows. The company gives a circle a bonus pool, and the circle lead decides how to split it — even how much to give himself — while HR and functional departments only have an advisory voice. Chen Bin says at this stage the ‘fighting’ line dominates; the functional line may in future become something like a technology committee, but within the first half year to a year of the transformation the front line gets to decide.
— Chen BinExperience can become a liability
Chen Bin says AI management trainees naturally started using AI from their university years in 2023, have no fear about doing things, and use it more than the veterans. His exact words: ‘sometimes experience instead becomes a burden’, and the most troublesome are the half-baked — whose experience hasn't arrived but who feel they are very experienced and can't change. Veterans have framework thinking and systematized thinking as their foundation, but the clear-eyed ones find it easier to judge which processes and which roles can be removed. The bar for AI management trainees is internal AI proficiency L2 or above, requiring skills and Agent-building ability; only L3 can author workflows.
— Chen BinR&D turns fast, marketing turns slow
Chen Bin admits transformation speed varies a lot by department: technical R&D tied to software and coding is easiest, because writing code every day gives positive feedback; marketing and branding are slow, because after publishing more content they ask ‘so what’ — did it bring revenue growth and bigger bonuses. He also mentions a real conflict: people doing design and content feel AI can only accelerate volume, not raise quality, and that 60-point output doesn't stick around. His response is to give everyone an operating mindset: whether content can drive conversion is the standard.
— Chen BinThree rounds of sign-ups: stand up again, then sit down again
Lockin ran three internal rounds of sign-ups: the first for incremental roles, such as a dozen or two company-level Agents and the corresponding Agent Builders and FDEs; the second for roles missing after circles formed, such as a content marketing Agent Builder; the third to keep filling gaps. Applicants first do a self-assessment to find what they're good at, then give a three-to-five-minute defense, and after the judges score, AI is used to analyze the meeting minutes and the talks for matching. Of a hundred applicants, twenty or thirty may be chosen in the end. Chen Bin says both the mechanism and the results were very good.
— Chen BinTo hire, you must first prove AI can't do the job
After a circle picks over the people in an original department, if the boss wants to add headcount there are logically only two channels: find an AI management trainee, or use more AI tokens. If you really must hire again, it goes to Chen Bin for special approval, and you have to explain clearly why a management trainee won't work and why AI won't work. Chen Bin says ‘logically he has no channel left’. They plan to hire 50 AI management trainees by year-end and are already close to 20. Xiaofei's role at Lockin is like a strategic HR or CHO as a service, working with Chen Bin on the AI-native transformation.
— Chen BinIPD isn't overturned, only accelerated
Chen Bin says inside the company he is in fact the most radical one, and people often pull him back to be more cautious. On Huawei IPD and similar accumulated practice, his judgment is: AI can only rework processes that were already unreasonable and especially redundant; IPD itself is very reasonable and can only be accelerated. For example, going from strategy to execution, SP to BP, used to take three months to half a year, now two months — double the efficiency — but the nodes like Five Looks Three Sets are still there, just three days instead of three weeks. Training a product line manager used to take 10 to 15 years, now three to five, but the rotation posts that must be done cannot be skipped.
— Chen BinNot every company should run this fast
Chen Bin's advice to three kinds of bosses — milk tea shops, PCB factories and call centers — is to look at the real value AI brings in that industry: the more traditional the industry, the lower the value, so you can wait until platforms like Doubao and Volcano Engine mature and then choose, with lower trial-and-error cost. But the more frontier the tech company, the earlier it should embrace AI, because technology changes a lot and the market changes fast, so decision quality matters more. He specifically notes that before physical AI takes off, PCB factories and the like should be cautious and not move first; milk tea shops are more channel- and operations-driven, and AI can quickly build a feedback loop from the user end to the supply chain end, so it matters more there.
— Chen BinIn their own words · checked verbatim
I think an AI-native organization probably starts from AI first — whatever AI can do, try not to have a person do it. Think with that AI-native mindset first, and think about this thing, even this process, as something that logically should be rebuilt.
就是我觉得原生组织 可能首先还是从AI出发 就是什么东西能让AI做 就尽量不要让人做 就先AI native的思考 去思考这个事情 甚至这个流程 逻辑上都应该被重构掉
Chen Bin4:01
It used to be functional, right? Now I'm an end-to-end small team around a value loop, and I even deliberately erase the functional power, or the boundaries — make functional power smaller and the circle's power, or the circle lead's power, bigger.
就是原来是职能型的 对吧 现在我就是端到端的 价值必还的小团队 甚至我就 就故意把职能的权力 让他要 或者边界 或者要把它给抹去 让职能的权力变小 让圈子的权力 或者圈长的权力变大
Chen Bin16:06
Actually sometimes experience instead becomes a burden. Better to have it, but the half-baked ones are scary — their experience hasn't arrived yet, but they feel they're already very experienced, right? Can't change.
其实有的时候经验反而会成为一个累赘 有更好 往往是那种半通水就比较恐怖 就是他经验还没有到 但他觉得已经很有经验了 对吧 不能改
Chen Bin21:07
As far as possible we fill with AI, right? Prioritize AI — whatever AI can fill, let AI fill it. Unless it can't be filled. And if it can't be filled, internally we say: either you say you'll find an AI management trainee to fill it, or you use more AI tokens.
尽量我们是以AI来补嘛 对吧 优先用AI镜头 用AI能够补的就AI来补 除非是补不了 那补不了呢 就我们现在内部 我们内部是说 要么你说找AI管培生来补 要么就是用AI更多的token
Chen Bin41:16
It used to take at least three months to half a year; now we do SP to BP in two months, right? At least double the efficiency. But the process itself hasn't changed — it's just that where I used to hold five meetings, now it's three meetings, two meetings.
原来可能至少花三个月到半年 那现在我们两个月 我就把SP给BP 对吧 就至少提效了一倍 但是它的流程本身没有改变 只是说呢 我原来开五次会 变成开三次会 开两次会
Chen Bin53:22
Actually look at the real value AI brings in this industry. I think the more traditional the industry, the lower the value it brings — lower, so you can reasonably say you can wait until these — people like us who are willing to fall in the pit and play around, right? Or things like Doubao at work, Volcano Engine — until those are done well, then you just choose.
其实看AI在这个行业里面 它的真正的带来的价值 我觉得越是传统的行业 它可能带来价值 确实它会越低一点 越低的话 那你就可以mari的说 你可以等到这些 像我们这种贪坑的贪玩的对吧 或者说像什么豆包工作 火山引擎 这些做的比较好的之后 你再去选择就好
Chen Bin1:00:24
Figures
| Years since Lockin was founded | 12 years | 2:00 |
| Share of Xiaomi smart locks developed and manufactured by Lockin | about 70% | 1:00 |
| Year-end hiring target for AI management trainees | 50 | 42:16 |
| AI management trainees already hired | close to 20 | 42:16 |
| SP to BP, previous time | three months to half a year | 53:22 |
| SP to BP, current time | two months | 53:22 |
| Training a product line manager, previous time | 10 to 15 years | 55:22 |
| Training a product line manager, current time | three to five years | 55:22 |
| Huawei rotation cycle | three years per post | 55:22 |
| Lockin rotation cycle | one year, even half a year per post | 55:22 |
Glossary
- HARO / Human AI Resource Office
- The department Lockin formed by merging HR and IT, responsible for AI transformation and org restructuring.
- FDE / Forward Deployed Engineer
- A role that goes into the business front line to map processes and land AI in the business; Lockin requires it to be a business expert.
- Agent Builder
- A technical-leaning role responsible for building Agents; short term it splits from the FDE, long term the two may merge.
- AI Buff
- Lockin's measure of the efficiency gain AI brings inside a circle or role, expressed as a percentage.
- IPD / Integrated Product Development
- Huawei's product development process system, which Chen Bin believes AI can only accelerate, not overturn.
- SP / BP / Strategy Plan / Business Plan
- The two stages from strategy to execution in Huawei's DSTE process, whose cycle Lockin compressed with AI.
How to listen
Founders driving AI org transformation, heads of HR and business leads — especially at hardware or consumer electronics companies who want to move on org design but don't know where to start.
After 1:04:25, the chit-chat and outro music can be skipped.