Outsourcing AI short video in Malaysia: 7 questions to ask before you pay anyone

The AI video market in Malaysia got crowded fast. Some of it is very good. A lot of it is a landing page, a subscription to a generator, and a promise of fast results.

I am not going to pretend I am a neutral party here. We produce AI short video for brands, so I have an interest in how you choose. What I can do is tell you which questions actually reveal something, including the ones that are uncomfortable for us to answer.

Start by deciding what you are buying

There are two products hiding under the same label.

The first is output. You pay, you receive a set of finished videos, the relationship ends. That is a legitimate purchase. If you need five clips for a launch next month, this is what you want, and a long engagement would be a waste of your money.

The second is a content operation. Someone is responsible for your brand having something worth watching in the feed every week, for reading what worked, and for changing the next batch accordingly. That is a different service with a different shape and a different price.

Confusing the two is where most bad outsourcing decisions start. A brand buys output, expects an operation, and concludes six weeks later that AI video does not work.

The seven questions

Here they are in short form, with the standard a solid answer should meet. Use this list on every vendor you speak to, including us. The rest of this article explains what each question is really testing.

One note on reading the answers. Replies that stay general, such as "leave everything to us", "unlimited revisions" or "AI works for any product", are not automatically a bad sign. Often they just mean the process behind them has never been written down. Press once for the specifics. How a vendor responds to being asked twice tells you more than the first answer did.

1Who writes the script, and do I see it first?

Short video lives or dies on the first three seconds, and the first three seconds are a writing problem before they are a production problem. If a vendor cannot show you the thinking before the rendering, they are decorating rather than selling.

Our own sequence is to understand the brand and the product being pushed, look at competitor work and any reference videos the brand likes, break down why those work, then write the script. The brand sees the script, or a first finished video, before we produce at volume. What you should not accept is a vendor who cannot articulate why a video opens the way it does. See how the first three seconds decide distribution for the mechanics behind this.

2What do you know about my selling platform?

A beautiful video that ignores how Shopee, Lazada or TikTok Shop actually distribute content is an expensive screensaver. TikTok Shop, for example, applies a content posting limit to seller accounts that publish repeated non-interactive videos, which means volume without quality can actively work against you.^[1]

You are testing whether the vendor thinks about conversion and platform mechanics, or only about aesthetics. Both matter. Only one of them is commonly missing.

3How many revisions, and when do they happen?

"Unlimited revisions" sounds generous and usually signals that nobody has thought about process. Endless revision cycles are how projects stall.

The structure that works is front-loaded. Get the tone right once, in detail, on the first piece, then produce the rest against that locked standard. We allow up to three rounds of revision on the first video, because that is where tone, pacing and brand feel are actually decided. Once it is agreed, the rest of the month follows the same rules and rarely needs rework. What matters is not the number. It is whether the vendor can explain where in the process revisions belong.

4Where do the finished files live, and what may I do with them?

This one is quietly the most valuable, and almost nobody asks it. A month of content is an asset you should be able to use in more than one place.

We upload finished videos to the client's own cloud storage, and the brand decides what happens next: posting across their own channels, handing the files to dealers and distributors, or reusing them in later campaigns. If a vendor keeps your content locked inside their platform, or cannot answer the question clearly, you are renting rather than building. There is more on this in why your videos should be treated as brand assets.

5How will we know which videos worked?

Ask what happens after publishing. A vendor who is accountable for outcomes will describe some form of measurement and elimination: run a batch, put a small amount of ad spend behind selected videos, watch three-second retention and interaction, then drop the weakest formats and scale the strongest.

If the answer stops at "we deliver the videos", that is your answer to question one about what you are buying.

6Which products would you tell me not to use AI for?

This is the honesty test, and it is the one I would ask first if I were the buyer.

AI content is not the right lead format for everything. Categories that run on personal trust and sensory reassurance, such as skincare, supplements and baby care, still need a real human face at the decision-making moment. So does founder-led storytelling and anything where the buyer is really asking "can I trust the person behind this". A vendor who says AI works for everything is either inexperienced or selling. Our own view of where the line sits is set out in which products suit AI video and which still need a real person.

7Who is actually producing this, and is any of it subcontracted?

Malaysian content has to work in English, Bahasa Melayu and Chinese, and it has to sound like it was made by someone who lives here. Humour, references and buying triggers do not survive being routed through a production line that has never sold into this market.

Ask plainly whether the team is local and whether any part of the work is passed on. We produce in-house with a local team and do not subcontract, which is the only reason we can hold tone consistent across three languages.

The contract question, answered honestly

You will read advice telling you to avoid long commitments. I want to address that directly, because we ask clients for a one-year agreement and you deserve to know why before you talk to us.

Contract length is really a question about accountability. A vendor who delivers files can work month to month, because their responsibility ends at handover. A vendor accountable for performance needs enough runway to complete the loop: publish, read the numbers, adjust, publish again. Our first month establishes the baseline. From there we review what the data says about completion, interaction and conversion, and the following month is built on that reading rather than on guesswork.

There is also a switching cost that brands routinely underestimate. Changing content partners every few months means re-explaining the brand from zero, absorbing a visible shift in tone, and throwing away everything learned about what your audience responds to.

What the market charges, and what you are really deciding

Published 2026 rates are worth knowing before anyone quotes you, mostly so you can tell what a quote is actually buying.

Notice what those numbers are priced against. They are all priced per piece of work, because traditional production charges you per shoot. Your tenth video costs roughly what your first one did, so the question quietly becomes how few videos you can get away with.

That is the real decision, and it is not a pricing decision. It is whether your brand can hold a presence in the feed every week for a year, or whether you are buying a handful of videos and hoping they carry you. Those two outcomes are not different prices for the same thing. They are different things.

Hansen's ad-grade AI short video is a fixed monthly engagement starting from RM2,990, with volume scoped to your brand on a strategy call. We do not position that as the low-cost option, and if you are choosing on price alone you should not choose us. The three routes are compared properly in what short video really costs to produce in Malaysia.

One last filter, and it is the harshest one

If a vendor leads with cheap and fast, walk.

It is 2026. Every brand in your category has access to the same tools you do. If results could be bought quickly and cheaply, your competitors would have bought them already and the advantage would have evaporated the same week. Anyone still selling that promise knows this. They are not confused. They are counting on you not doing the arithmetic.

But it takes two. This end of the market survives because there is always a buyer who wants to believe it, who would rather pay a little and be told what they want to hear than pay properly and be told the truth. That buyer is not really being deceived. They are volunteering, and then calling it bad luck when the invoice produces nothing.

The brands that win in the feed did not find a shortcut. They showed up, week after week, with something worth watching, for long enough that it started to compound. That is unglamorous, it is slow at the start, and it is the only version of this that has ever worked.