Is the spirit of craftsmanship still needed in the AI era?
In the era of artificial intelligence, two modes of production are unfolding simultaneously.
A group of entrepreneurs use large models to complete market research, write code, and develop product prototypes within a few days, while advancing multiple directions at the same time, letting user feedback decide which solutions can survive. The other group still sticks to one dish, one precision tool, one bearing, spending years narrowing the error by a fraction of a millimeter and slightly elevating the user experience.
The former believes in speed, while the latter believes in precision. Over the past few decades, the precision-focused camp has almost monopolized the moral high ground, and their traits of patience, concentration, and the pursuit of excellence have been taken as the default standard for great enterprises. This narrative works very well in a stable industrial structure, but it hides an assumption that has never been seriously tested: the direction itself is correct.
Once the direction changes, all the patience needs to be revalued. The harder a person works on the wrong track, the further away they get from the market, and the time they consume and the experience they accumulate will instead become liabilities. The craftsmanship spirit itself has not expired, but it cannot be evaluated independently without considering the industrial direction, technology cycle and market demand. What we really need to ask today is not whether we should continue to pursue excellence, but where we should apply this pursuit of excellence.
The Disappearing Products
Most of the technological progress in the industrial era took place within the same product framework. A car is still a car, but the engine efficiency has improved; a camera is still a camera, but the lens is clearer; a TV is still a TV, but the screen is larger and more vivid. This world provides the best soil for the craftsmanship spirit. The product life cycle is long enough, and the skills practiced today can still be used 20 years later, so enterprises are willing to spend ten years cultivating a senior technician.
Artificial intelligence does not stop at the level of "making tools easier to use". It directly rewrites whether the product itself needs to exist. The functions of a set of software become an instruction in a large model; the service that can only be provided by a professional team becomes a natural language conversation. The opponents many enterprises encounter are no longer competitors that make similar products more exquisite, but customers who no longer need such products at all.
Quality cannot exist independently of direction at this time. The keys of a mechanical typewriter can be made extremely smooth, the color of film can be adjusted to extremely delicate, and the pistons and gearboxes of fuel engines can also be made extremely precise. But once the market shifts to digital documents, smartphones and electric vehicles, the stage for these precisions to play their role will shrink rapidly. Nokia's body is strong enough, Sony Walkman was once the ceiling of portable music, and Kodak holds the best film technology in the world. These enterprises are not short of quality or engineers, but lack the awareness to admit in time that the entire product system has changed its foundation. The real reason why many enterprises die is that they have solved a problem that is losing significance extremely beautifully.
The part of the craftsmanship spirit that is most prone to problems is precisely the part that is most praised. The deeper the investment, the heavier the emotion, and the harder it is to turn back. A team that spends 20 years polishing a technology will naturally believe that it still has a future. Admitting that the technology is obsolete means that all the equipment, processes and professional identities accumulated over these years have to be revalued, and no one is willing to take the initiative to bear this loss.
As a result, the pursuit of excellence has quietly become a shield to protect sunk costs. Enterprises continue to polish a product that has no room for growth, and take "adhering to quality" as a reason for unwillingness to change. The real mature craftsmanship spirit, in addition to the ability to do one thing well, also has the vision to judge whether this thing is still worth doing. In a stable era, working hard will bring returns; in an era of drastic technological changes, you need to look up at the road before you bow down to work.
MVP > Perfection
The MVP mentioned in the entrepreneurial circle refers to the Minimum Viable Product. It does not require enterprises to deliver mature works at the very beginning, but only to develop the most basic functions, present them to real users, and verify the most critical business assumption.
Traditional product development is a straight line: plan first, then research and development, and then launch to the market after long-term testing. There is an assumption behind this that enterprises have a considerable grasp of user needs, and the product will not be completely overthrown after launch. MVP adopts a different strategy: first make a prototype at low cost, see the market response, and then decide whether to increase investment. What it really cares about is whether the enterprise can find out its mistakes as soon as possible, and whether the first version is exquisite enough is a secondary issue.
Artificial intelligence has reduced the cost of trial and error to an extremely low level, and has also made the MVP methodology more effective. In the past, to make a software product, product managers, programmers, designers, testers and marketing teams had to work together for several months; today, a skilled entrepreneur with large models, automatic programming tools and ready-made cloud services can piece together the first available version in a few days.
After the development cost is reduced, there is no need for enterprises to bet all their resources on a carefully deduced plan. They can advance multiple directions at the same time, and let real users rather than internal meetings vote. The real reason why many products die is that no one needs them at all, and factors such as whether the buttons look good, whether the speed is fast, and whether the functions are complete are not that important. When an enterprise spends a year polishing details, it is only delaying the time to hear the market say "no".
The window of opportunity in the AI era is also shortening. A product idea that is still fresh today may become a standard function of a general large model a few months later. The more an enterprise wants to make the first version complete, the more likely it is to find that the market has already moved on the day of release. Therefore, in fields with rapid technological changes, unclear demand and controllable trial and error cost, the value of MVP usually surpasses the traditional craftsmanship spirit. Verify the direction first, then invest heavily; prove that someone is willing to use the product first, then discuss how to make it perfect.
This does not mean that MVP can be perfunctory. Users can accept fewer functions, but cannot accept fake functions; they can accept a rough interface, but cannot accept data loss without warning; they can accept that the product is still in the testing phase, but cannot accept that the enterprise transfers risks to them secretly. A good MVP subtracts from functions and adds to core value: it cuts off the decorations that are not needed for the time being, and retains honesty, reliability and basic quality.
The craftsmanship spirit and MVP are not opposites. They are two stages in the product life cycle. MVP is responsible for finding the direction and answering whether there is demand for the product; the craftsmanship spirit is responsible for building a moat and answering whether others can copy it easily. Before the direction is verified, the more polishing you do, the more waste you will have; after the direction is verified, insufficient polishing will make it difficult to form a long-term barrier. The product logic that works in the AI era is usually to rush out first, and then approach a better answer round by round — you usually can't wait for the day when the product is perfect if you try to achieve perfection at one go.
The Slow Transformation of Germany and Japan
Germany and Japan are often regarded as two models of the craftsmanship spirit.
Germany has a large number of small and medium-sized enterprises deeply engaged in niche markets. They may only produce one type of pump, one type of valve, one type of bearing or sensor, but they can occupy an important share of the global market. They do not rely on traffic or chase hot spots, but build thresholds through decades of technology accumulation, customer relationships and engineering experience. This model supports the basic framework of Germany's manufacturing industry.
Concentration brings accumulation, but also brings path dependence. The foundation of Germany's automotive industry is mechanical engineering, precision processing, engines, gearboxes and a complete system of complex parts. A fuel vehicle supports a huge supply chain, and every link can support a professional company that has been deeply engaged for many years. Electric vehicles have removed most of this foundation. The power system has shifted from complex mechanical devices to batteries, motors, power semiconductors and software. The problem is no longer whether the processing precision is sufficient, but that vehicle manufacturers no longer purchase such parts at all. No matter how precise the fuel engine parts are, they cannot change the fact that electric vehicles do not need them.
The pressure on Germany's automotive industry in recent years certainly cannot be entirely attributed to the craftsmanship spirit. Energy prices, labor costs, regulatory burdens and international competition are all playing a role. However, the long-term formed technological inertia does make many enterprises better at continuing to optimize within the old framework, but slower to switch the entire product system, and they tend to take past advantages as the law for the future. But electric vehicles and artificial intelligence are increasingly relying on software, data, chips and ecosystems. The basic unit of competition has expanded from a single part or a single piece of equipment to the entire platform. Details are still important, but the details that determine the outcome have changed their positions. In the past, they were engines and gearboxes; in the future, they will be operating systems, autonomous driving algorithms, chip architectures and data closed loops. Polishing old parts to the extreme cannot bring discourse power in the new system.
Japan's problems are different from Germany's, but there is a similar cultural inertia in essence. From sushi, knives, ceramics to automobiles, cameras and consumer electronics, Japanese society has long advocated "spending a lifetime doing one thing well". This culture can foster an extremely stable production process, and can also make many ordinary products have the texture of works of art. In industries with stable demand, long-term concentration is very valuable. What the master passes on to the apprentice is not only technical parameters, but also implicit experience that cannot be written into the manual, such as the subtle changes of materials at specific temperatures.
Japanese enterprises once built excellent reputation in the fields of televisions, Walkmans, digital cameras and home appliances. They are good at improving hardware, adding functions and reducing failure rates, but they lost their dominant position in the competition of internet platforms, mobile operating systems and software ecosystems. There are multiple reasons behind this, including corporate governance, capital market and demographic structure. However, over-reliance on existing processes has indeed weakened the organization's ability to self-deny. Every part has responsible personnel, but the entire system is increasingly unable to keep up with the new competition rhythm.
The sushi story of Jiro Ono can be regarded as a symbol of personal craftsmanship, but it cannot be applied to the competition logic of the technology industry. A restaurant can spend decades polishing a flavor, because what customers buy is this time and concentration; the technology market will not wait that long. Consumers may replace a tool within a few months, and the platform may rewrite the rules overnight.
The real dilemma of Germany and Japan is how to transfer their accumulated precision manufacturing capabilities and long-term experience to a brand new industrial structure, which has little to do with having too many craftsmen or too good products. Once the craftsmanship spirit is understood as sticking to old technologies and rejecting rapid trial and error, it will change from a competitive advantage to a cultural burden.
What they should retain is concentration, quality and sense of responsibility, and what they should discard is the superstition of a specific product or a set of old processes.
Tracks That Require Extreme Efforts
If you only see the impact on traditional industries, it is easy to draw a lazy conclusion: in the era of artificial intelligence, we only need speed and do not need to pursue perfection anymore.
The reality is completely different. The further artificial intelligence develops, the higher the requirements for the stability of computing power, chips, manufacturing equipment and materials. The faster the algorithm iterates, the higher the precision threshold of underlying physical products.
Japanese enterprises still hold strong positions in the field of semiconductor materials. Companies such as Shin-Etsu Chemical, Tokyo Ohka Kogyo and JSR have been deeply engaged in silicon wafers, photoresists and high-purity chemical materials for decades. Their products look unremarkable, but they directly determine the precision, yield and stability of chip manufacturing. A tiny extra impurity in the material may scrap the entire batch of wafers. This capability cannot be achieved in a hurry, and it is accumulated through decades of experimental data, equipment experience and customer verification.
Ajinomoto is another example. This company, which started as a seasoning manufacturer, redirected its accumulated amino acid chemical technology and developed ABF insulating films for high-performance chip packaging. The faster AI servers, GPUs and advanced chips run, the more such seemingly marginal materials become key bottlenecks.
Zeiss from Germany also shows that focusing on a narrow field does not mean falling behind. Zeiss has been deeply engaged in high-end optics for decades, supplying core optical systems for advanced lithography equipment. The surface error of the lens needs to be controlled at an extremely tiny level. This kind of product cannot be "launched first and then iterated". A tiny error may collapse the entire equipment and the entire chip production line.
These enterprises are also concentrated in niche fields. The difference is that they are stuck at the bottleneck positions of the new industry. Artificial intelligence will not weaken the value of such products, but will only amplify the demand for them. The higher the model parameters are stacked, the more advanced the chip process is, and all of these will eventually fall to the three things: material purity, equipment precision and manufacturing yield.
This shows that to judge whether the craftsmanship spirit is valuable, we cannot only look at whether the enterprise operates in a narrow field, but also look at its position in the industrial chain. A traditional engine part can be the best in the world, but the entire technical route is shrinking; a chip insulating film looks unremarkable, but it may be an indispensable material for advanced packaging. The track determines demand, the bottleneck determines profit, and the craftsmanship spirit is responsible for turning the bottleneck into a moat.
In addition to semiconductor materials and high-end equipment, fields such as aero-engines, medical devices, nuclear power equipment and food safety also cannot blindly believe in MVP. When human life, safety and high-value assets are involved, products cannot be launched with a rough version in the real environment and then modified after an accident. Luxury goods, high-end catering, artworks and traditional handicrafts belong to another category. What consumers buy is not only the function, but also the time, craftsmanship and story. Artificial intelligence can imitate the appearance, but it is difficult to replace the cultural value brought by "someone has spent many years on it".
The craftsmanship spirit has not disappeared, and its high-value interval is being redefined. In places with highly uncertain demand and low trial and error cost, speed and MVP are more valuable; in places where the cost of error is extremely high, or the product value is supported by extreme precision, long-term trust and non-replicable accumulation, the craftsmanship spirit is still irreplaceable. The key of the problem is which link should be fast and which link must be slow. Speed and slowness are not inherently superior to each other.
The New Craftsmanship Spirit
Artificial intelligence is not the opponent of the craftsmanship spirit. The large model itself is an extremely complex engineering product. From data cleaning, algorithm design, chip clusters to model training, inference optimization and safety testing, every link relies on long-term accumulation and a large number of details.
The biggest difference between AI enterprises and traditional craftsmen is that they do not take a certain generation of products as the end point. The model is continuously updated, and errors are corrected through feedback. What they pursue is the ability of continuous evolution, rather than achieving perfection at one go.
This may be the most important change of the craftsmanship spirit today. In the past, the craftsmanship spirit relied on repetition, spending a lifetime polishing the same thing to make it better; today's craftsmanship spirit also needs to add the capabilities of choice, migration and self-denial. A person should not only know how to make products better, but also know when to change tools, change methods, and even replace the product he is producing. The real long-termism never means sticking to a specific practice. What enterprises should pursue in the long run is to solve problems, create value and build capabilities. The product form is just a phased carrier.
The most dangerous craftsmen in the AI era are those who regard their skills as their identity and the path they take as their belief. Such people may have extremely high capabilities, but they refuse to admit that the market has raised new problems.
A more product logic that fits the current era probably has three steps. First, confirm the direction with MVP, and verify as early as possible whether the product can solve real problems, whether users are willing to use it, and whether the business model is feasible. Then concentrate resources on the links that really determine the outcome. Not every button or every part is worth extreme efforts. Enterprises need to find the part that users care most about and that competitors are hardest to copy, and invest limited capital and manpower there. Finally, retain the ability to adjust direction at any time. Even if the product is successful, you cannot take the current form as the permanent answer.
To put it bluntly, the ability to judge direction determines whether your efforts are effective, the iteration speed determines whether the enterprise can seize the window of opportunity, and the execution precision determines whether the product can build a moat. The craftsmanship spirit only solves the problem of execution precision. Without direction judgment, the higher the precision, the greater the waste; without iteration speed, the opportunity will disappear before the product is completed; without execution precision, even if you find the right direction, you can only stay at a level that is easy to be copied.
The truly competitive enterprises in the AI era do