# Why do AI productivity tools always start with professionals? A shortcut, or an illusion? - Date: 2026.01.12 - Locale: en - Page: https://cv.rocorgi.wang/writings/professional-or-illusion - Author: Long Wang (汪龙, rocorgi) Observations from building AI design tools for the China market: professionals can define correctness, but they are not necessarily how you scale. Some recent observations from building AI products, limited to the China market. Design tools are the example. Other tools may differ. Producers are always the minority. Most people do not have a strong motive to create. One fact is easy to miss: people accept AI as something to consume, and that acceptance does not transfer to AI as something to produce with. What you can observe: ordinary people will scroll AI videos, look at AI images, and forward AI jokes. Ask them to finish a real job with an AI tool — a poster they would actually post for an event — and the willingness drops. Many try once and leave. Not because it stopped being novel, but because they cannot produce something they dare to publish. This is not “users being conservative.” The consumption path and the production path are asymmetric. The place where content is consumed (short video, social) already supplies a standard: funny, good-looking, satisfying, or strange is enough. The place where content is produced asks you to own the result: usable, publishable, convertible, deliverable. A product that only lets someone “generate something that looks fine” is closer to a content toy. A product that lets someone deliver a result and live with the consequences has entered the productivity category. That is the break. The consumption world allows random surprise. The production world needs something stable and controllable. Many teams say ordinary users cannot write prompts, so the tool cannot scale. For most people the input really is plain speech, a casual remark, an unclear goal. If you treat “they cannot write prompts” as the core reason, you usually misread the problem. First, what ordinary users lack is not the ability to express themselves. It is an imagination of a good result — a standard of judgment. They are not unable to speak. They do not know what counts as good. When making a poster they say “make it more premium,” but in design “premium” can mean whitespace, contrast, hierarchy, type, brand consistency, control of information density. If the user has no picture of the result, every generation becomes a matter of luck. Luck means it cannot be reused and cannot be expected. It is hard to retain someone on that. Second, what they lack is not an understanding of AI. It is a place in the workflow: which step does this tool actually remove? For many people, content is not produced in “design software.” It is produced inside a platform: Douyin templates, CapCut, Xiaohongshu notes, a merchant backend. Asking them to open a new tool adds a branch outside that path. If the step does not pay off immediately — faster publishing, higher conversion, less rework — they go back to the old path. So the inability to write a prompt is the surface. What it exposes is a missing expectation of the result, plus a missing place in the workflow. From a product and business point of view, “start with professionals” is a reasonable strategy. It solves several early problems by default: Those benefits are real. The problem is that they also tempt you to ignore something else: professionals may be the hardest group to scale. Migration cost, asset lock-in, and the trap of catching up. Still using AI design tools as the example. When you actually go after professionals, three problems show up. Professionals are a good place to verify capability and to set a standard of correctness. They are not necessarily how you get to scale. Clear ROI, high frequency, good enough is enough. Separate professionals from power users and you notice how often teams mix the two. In the China market, the group that can scale earlier is often the power user. Still using design tools, typical profiles include: What they share: a clear ROI (efficiency, conversion, followers), high frequency, no deep lock-in to legacy assets, and a high tolerance for “good enough, and fast.” These people do not need you to replace the Adobe suite — the size of that Adobe population still surprises me — they need you to cut the path from idea to publish roughly in half. That is the more realistic path to scale. If you bet scale on teaching users to write prompts, you are taking on an education cost that is very hard to earn back. The more effective move is to let plain speech be a valid input, and let the system do the structuring. Mechanisms that can actually be built: The point is not a better chat. It is productizing a foreseeable result, plus constraints, plus delivery. An AI tool is not there to teach someone how to talk to AI. It is there so they can finish the job without having to learn. It is easy, when building AI tools, to fall into the illusion that “knowing how to use AI” is the goal. What the market actually rewards is getting the thing done. So is “start with professionals” a shortcut? My conclusion: if you treat professionals as the standard for judging results, and as a source of high-quality feedback, it is a shortcut. If you treat professionals as the audience you will scale with, it may be an illusion. The more realistic path: use professionals to define what a correct result is, use power users to scale, and let the system grow the prompt for the user. First, define the power user clearly. Their strength is ROI, not tool skill. Many teams imagine a power user as “semi-professional.” In China the more common picture is: strong demand, high frequency, accountable for the result (it has to go out today, and it has to convert), but unable to state design constraints, unclear on the limits of AI, and unwilling to learn a new paradigm. You cannot expect them to learn prompts. You have to turn their plain speech into structured input, so the output can be stable. Lowering that threshold matters. Second, if you worry that demand will make the tool complex, the root is that you are taking in capability requests. Users often say: Those are capability requests. They naturally drag the product toward catching up with the old tool, and the product becomes extremely complex. Translate the request into a result: ship faster, rework less, stay consistent, convert better, reuse more easily. Keep aligning on result metrics, and the complexity comes back down. These observations are only about AI design tools. ![Why do AI productivity tools always start with professionals? A shortcut, or an illusion?](/writings/professional-or-illusion/photo.jpg) [Original post](https://x.com/wangdouglas/status/2010661225622069525)