People ask me what I actually use AI for. The honest answer is that it depends on which hat I am wearing that day. Wearing the factory-owner hat, I use it for things that must get done before six o'clock. Wearing the entrepreneur hat, I use it to build things that do not exist yet. Wearing no hat at all — just a person at home — I use it for the same ordinary questions anyone else has.

Same tool. Three very different jobs. It is easier to show what I mean than to define it, so here is the whole list.

1. As a factory owner: what earns its keep every day

This is where AI has paid for itself the fastest, because a factory has a hundred small recurring tasks that are too small to hire for and too frequent to ignore.

• Image work is the most used, by far. Recolouring a bag into a new colourway, turning a plain sample photo into a clean product image, putting a mockup together for a customer — this used to wait for a designer or a photographer. Now most of it takes minutes.
• Copy, short-video scripts and the website. Product descriptions, promotional copy, scripts for short videos — and the company website itself. I never studied code; the site was written and is still maintained by describing what I want.
• Carton shipping marks. We have a thermal label printer. AI produced the label template, and now we type the PO number, quantity and size on a simple form and print directly. No more editing a Word file every time.
• A freight-quoting agent. Two rate spreadsheets went into the AI, and out came a tool that calculates the drop-ship courier fee for each customer. When there are thousands of orders, nobody looks them up by hand any more.
• Translating and answering overseas customers. Messages arrive in languages I do not read. Doubao or Qwen translates them; I tell the AI what tone to use and what we can and cannot promise, and it drafts the reply. The judgement stays mine, the typing does not.
• Product development. I describe a canvas tote idea to Doubao and get design directions back. Then we sample it, lock the artwork, and put it into production. The design conversation that used to need a meeting now happens in an afternoon.
• Invoice preparation. E-commerce invoice exports cannot be issued directly. We used to open one window, copy, switch, paste — for rows and rows. I had AI build a small tool that rearranges them instead.
• The full invoicing loop. Export from the platform, run it through the tool, issue in the tax system, download, re-import, and generate the platform's template. Several hundred rows now takes about ten minutes.
• Website upkeep. New factory photos, new product categories, small layout fixes. With no coding background, I make the change by describing it.

Next in line: finance, people management, payroll, approvals, customer records, material inventory. The plan is a small internal workspace rather than a stack of separate spreadsheets. I have looked at open-source options and built a piece of it, but this one is still early.

2. As an entrepreneur: what I am trying to build

The factory is the root. E-commerce is the cash flow that keeps the root watered. AI tools and platforms are where I think the next ten years go — not instead of the factory, but growing out of it.

What already exists: the website, a mini program, some software, and a toolbox — the invoicing tool and the freight agent above are part of it. Each one started as a chore I was tired of doing by hand.

One experiment I like more than it probably deserves: old phones. A drawer of idle Android handsets became a dashcam, a fatigue monitor, a pet camera, and an item tracker. No extra hardware, everything runs locally, and nothing leaves the device. The whole appeal is that the hardware is already paid for and the privacy is structural rather than promised.

The bigger idea is a platform built on six pillars: AI social, a place for founders to compare notes, legacy notes for passing on experience, robot customisation, AI guardianship, and remote teams. I do not claim it is finished. I claim it is worth building while the factory keeps running.

Which is also why the services side stays small and concrete: custom AI tools, canvas bag customisation and export, and consulting for one-person companies and AI-enabled startups. All three are things I can do today with people I already work with.

The mindset is simple: do not drop the factory, and do not wait for the perfect moment either. Watch, walk, build, repeat.

3. As an ordinary person: the unglamorous half

Away from work, the usage is almost boring, and that may be the most useful part of this article.

• I think out loud with AI — work problems, business ideas, small decisions I would otherwise turn over in my head at midnight.
• Photos, short videos and copy, the same as at work.
• Lyrics and poems, and turning them into songs. No client, no deadline, just something I wanted to try.
• Tutorials: how to do makeup properly, how to cook a dish I only know the name of.
• "What is this thing?" and everyday facts. That used to be a search engine and five tabs of ads; now it is one question.
• Food, housing, travel, what to wear. Small questions, answered immediately.

Closing: one tool, three angles

That is basically the whole list. The factory angle is about removing repetition. The entrepreneur angle is about turning what I learned removing repetition into something other people can use. The personal angle is about having a patient assistant for questions too small to bother anyone with.

What connects them is not intelligence. It is aim. AI is useful here because every single item above was a specific, annoying, recurring chore — not a vague wish to "use AI". When I could not name the chore, the results were worthless. When I could, the tool usually took an afternoon.

It still gets things wrong, and anything that leaves the factory is checked by a person before it goes out. But the direction is settled: in a traditional workshop like ours, AI is not a headline. It is a faster pair of hands.