After Execution Gets Cheap: How a Creative Studio Works with AI
The short version: AI has really changed only one thing. Execution got cheap. Once execution is cheap, the most expensive thing left is a point of view: what you want to make, why, and what counts as good. So our workflow at INTENTION (識聚有限公司, INTENTION CREATIVE CO., LTD.) is simple. The work that slows us down goes to tools. Judgment stays with people. Client files stay on our own computers.
Below is why I see it this way, what we actually do, and what I predict for the next two years.
1. What is happening now
Everyone is already using it. A 2025 survey of 16,000 creators found that 86% already use generative AI, most often for editing, upscaling, and retouching (55%), generating new assets (52%), and ideation (48%) [1].
Demand for video keeps getting pushed up. In a 2026 music industry survey, 75.3% of musicians said they feel pushed to keep producing video on top of their music, and 75.9% of respondents believe video directly affects whether their careers succeed or fail [2].
AI is eating the slowest part. Auto-dubbing is now routine: by late 2025, more than 6 million people a day were watching 10 minutes or more of auto-dubbed video [3]. Subtitles, transcripts, translation, and file conversion are all heading the same way.
But the time saved is smaller than you'd think. Research from the US Federal Reserve calculated that people who use generative AI save an average of 5.4% of their working hours, roughly 2.2 hours a week [4]. The tools are strong, but if the workflow doesn't change, that's all you save.
Everyone is afraid of data leaks, and everyone is uploading data anyway. 64% of people worry about accidentally leaking sensitive data, yet nearly half admit to putting non-public data into AI [5]. NDA clauses restricting AI use went from 0.65% in the first quarter of 2025 to 6.12% in the second quarter of 2026 [6].
2. How I see it
I only sort work into two kinds: things that slow you down, and things that need your judgment.
Things that slow you down, like writing up meetings, finding reference images, shrinking files enough to send, and turning speech into text, go to tools. The sooner the better. They have no point of view, only a time cost.
Things that need your judgment, like which direction to shoot, which version to pick, whether a frame is right, and how an artist wants to be seen, I don't hand off. AI can give me a hundred versions. Choosing one is the job of the director and the creative director.
The reasoning goes like this:
The cheaper execution gets, the more you can make with the same budget (points 1 and 3 in section 1).
There is more and more stuff, but the people watching still only have 24 hours a day. So what's scarce is no longer content. It's a trusted position.
Trust comes from judgment: what this community cares about right now, how this brand wants to be understood. AI can write a hundred posts. The person who can tell which one works decides whether those hundred posts are worth anything.
So "people with a point of view" are worth more than "people who are fast," and "people who are fast" are worth more than "people with strong technical skills." That is the order we hire and take on work by, and after AI it will only get more obvious.
I often tell new team members: don't just keep chasing and copying. Understand what you're actually doing. Tools change every quarter. Judgment stays with you for life.
3. How one of our proposals actually gets made
Stage | Where it used to get stuck | What we do now |
|---|---|---|
Finding references | What you saved was a blurry thumbnail, unusable once projected | We use Fully, which we built ourselves: paste a link, and images and videos save at original size |
Meetings | Nobody wrote it up afterward, and nobody remembered who said they'd do what | Tap Hearby before the meeting and you have notes when it ends, every line timestamped. In a test, a 70-second recording took 9 seconds to transcribe and 20 seconds to organize |
Writing the proposal | I could say the idea out loud, but my typing couldn't keep up | We use Talky to type by talking |
Sending it | An eight-page portfolio was 72.7 MB, and email wouldn't send it | Slimly compresses it to 7.5 MB, and zoomed in it's still the same image |
Production | Scripts, visual concepts, and edits had to be tried one version at a time | AI runs a lot of directions first, then people choose; final frames are mostly live-action, shot by our production arm NOUS ASIA |
None of the tools in this table help us "think." All they do is give us our time back. We treat AI as an exploration tool, not a delivery tool.
The four tools have one more thing in common: your data doesn't leave your computer. That isn't a feature. It's basic courtesy to our clients (see point 5 in section 1).
4. Three things I deliberately don't hand to AI
Direction. First ask "what does this project really need," then ask what to make and how. Nobody can do this step for you.
Anything involving people. Casting, talking with artists and brands, judgment calls on set. You can't generate a relationship.
Final quality. If something is worth shooting, it's worth shooting well. Tools make us faster. They don't make us lower the bar.
5. My predictions
Our company keeps a prediction ledger. Each entry states its due date, its probability, and what counts as right, and we settle it when it comes due. Our in-house trend researcher estimates the probabilities, and I read them before deciding whether to bet that way. Three entries relate to this piece:
Cheap one-off assets will be absorbed directly by the platforms. One platform has publicly set a goal of letting brands supply only a product image and a budget, while AI makes and places the ads automatically [7]. Prediction: by the end of 2026, this feature opens to ordinary advertisers in Taiwan. The ledger puts it at 40%.
A Taiwanese record label will release its first official MV made mainly with AI. Prediction: by June 2027, at least one major record label releases an official MV whose main visuals are AI-generated and labeled as AI. The ledger puts it at 75%.
But work that is human, present, and credible will be worth more. Global recorded music revenue was US$31.7 billion in 2025, its 11th straight year of growth [8]; creator ad spend was about US$37 billion in 2025, up 26% year over year [9]; 45% of listeners are uncomfortable with AI writing original music [10]. Prediction: our number of new projects with artists and brands in the first half of 2027 will be no lower than in the first half of 2026. The ledger puts it at 55%.
If 1 and 2 come true, that's "execution" getting absorbed. If 3 comes true, it means "judgment" and "people" are worth even more. All three holding at once is the conclusion of this piece. When they come due, we'll come back here and write up the results.
6. Three suggestions for creators
Start by listing the three things that get you stuck most often each week. Don't look for the hottest tool. Look for the part that wastes your time most often.
Hand off what repeats, and keep the judgment for yourself. Understanding logic isn't the same as knowing how to code. It's three moves: break it down, describe it, check it. AI can only get as far as you can explain clearly.
Don't casually upload client files. Contracts, unreleased work, meeting recordings: as much as possible, use tools that handle them on your own computer.
The four tools we built are all on INTENTION Made, free and open source: https://www.intention.ltd/made
Sources
Adobe, Creators' Toolkit Report (16,000 people, 8 countries), 2025-10-28. https://news.adobe.com/news/2025/10/adobe-max-2025-creators-survey
Berklee Emerging Artistic Technology Lab, In Sync: Music and Video 2026 (over 1,000 people), 2026-06-04. https://www.berklee.edu/beatl/in-sync-music-and-video-2026
YouTube Blog, How YouTube auto dubbing helps creators reach global audiences, 2026-05-15. https://blog.youtube/creator-and-artist-stories/youtube-auto-dubbing-explained/
Federal Reserve Bank of St. Louis (Bick, Blandin, Deming), The Impact of Generative AI on Work Productivity, 2025-02-27. https://www.stlouisfed.org/on-the-economy/2025/feb/impact-generative-ai-work-productivity
Cisco, 2025 Data Privacy Benchmark Study (2,600 people, 12 countries), 2025-04-02. https://newsroom.cisco.com/c/r/newsroom/en/us/a/y2025/m04/cisco-2025-data-privacy-benchmark-study-privacy-landscape-grows-increasingly-complex-in-the-age-of-ai.html
Ontra, AI-use restrictions are rising, but still only appear in 6% of NDAs (645,000 NDAs), 2026-09-15. https://www.ontra.ai/blog/ai-use-restrictions-rising-ndas/
Reuters (via Yahoo Finance), Meta aims to fully automate advertising with AI, 2025-06-02. https://finance.yahoo.com/news/meta-aims-fully-automate-advertising-113640522.html
IFPI, Global Music Report 2026, 2026-03-18. https://www.ifpi.org/global-music-report-2026-global-recorded-music-revenues-grow-6-4-as-record-companies-drive-innovation/
IAB, 2025 Creator Economy Ad Spend & Strategy Report, 2025-11-20. https://www.iab.com/insights/2025-creator-economy-ad-spend-strategy-report/
Luminate, 2025 Year-End Music Report, 2026-01. https://luminatedata.com/reports/yearend-music-industry-report-2025/
About the author
陳韋傑 (Aloha jie) is the founder of INTENTION (識聚有限公司, INTENTION CREATIVE CO., LTD.) and a director. INTENTION is a creative strategy and visual production company in Taipei, founded in 2022, working on album visuals, music videos, brand films and social content. Production is handled by its production arm NOUS ASIA (冉冉創意股份有限公司). INTENTION Made is where the team turns the friction in its own creative workflow into free tools.
留言