Most weeks, at least one prospective client asks the same question in a different wrapper: could we just get an AI tool to make this instead of commissioning a studio? It is not an unreasonable thing to ask. The market backs it up: AI video generation is now an $847m industry growing at close to 19 percent a year, expanding roughly three times faster than the broader video editing software category, and 63 percent of video marketers say they have already folded some form of AI into their workflow.
We use these tools ourselves, in specific places, for specific reasons. This piece is about where that actually holds up in practice, where it falls apart for a fintech or SaaS brand film specifically, and what a run of AI-generated advertising over the past two years has told us about how audiences respond when a brand leans on it too visibly.
The numbers are real, but they measure the wrong thing
The economics are genuinely striking. A minute of AI-generated footage can cost somewhere between $2 and $30 to produce, against $1,000 to $50,000 a minute for a traditionally shot and animated equivalent. A 60-second video that used to take around 13 days can now be roughed out in under half an hour. Adoption is climbing fast too: 91 percent of businesses now use video for marketing in some form, and 45 percent of B2B marketers name AI-powered tools as their top investment priority for 2026, ahead of every other category by a clear margin.
None of that is in dispute. What it measures, though, is production speed and unit cost, not whether the finished film does the job a brand film is actually there to do: make a complicated product legible, and make the company behind it look like one worth trusting with money. Those are different problems, and the gap between them is where most of the disappointment happens.
Where AI tools already earn their place
We are not precious about this. AI tools have found a genuine home in the early and middle stages of the process, where the requirement is speed and volume rather than final polish.
Pre-visualisation is the clearest case. Generating a rough animatic to test pacing, or a handful of style frames to settle a colour direction before a single frame is properly rendered, used to eat days of a project's timeline. Now it takes an afternoon, and clients can react to something concrete instead of a mood board and a promise. Temp voiceover for internal review, background plates and textures that will never be the hero of a shot, quick variations on a graphic treatment to compare against each other: all of this has got faster and cheaper without costing anything in quality, because none of it was ever the thing the client was actually paying for.
There is a second category worth naming honestly: short, disposable, high-volume content. A weekly LinkedIn clip explaining a minor product update, a quick social cutdown, an internal training video. If the brief is genuinely "we need something serviceable by Thursday and nobody is going to watch it twice," an AI tool is often the right tool, and we will say so rather than pitch a five-figure production for it.
Where it comes apart for fintech and SaaS
Brand films and explainer videos in this world are carrying more weight than most video content has to. A fintech product is, by definition, asking someone to trust it with money, or with the infrastructure their own customers' money moves through. A SaaS platform is usually asking a buying committee to bet a budget line and their own credibility on a piece of software they may never fully see under the hood. The film's job is to close some of that trust gap. Anything in it that reads as slightly synthetic works against that job, even when the viewer cannot articulate why.
This is also where the "invisible product" problem, something we have written about before in a different context, bites hardest. Abstract financial infrastructure, tokenisation, settlement logic, anything without a physical form or an obvious on-screen interface, needs a level of deliberate, specific visual metaphor that current AI video tools are not built to sustain. Ask a general-purpose model for "data flowing securely between two systems" and it will give you something that looks plausible for four seconds and then starts to drift: proportions shift, a logo warps, a hand gains a sixth finger in the background. For a 30-second social clip that is a shrug. For the film that sits on your homepage answering the question "can we trust this company with our money," it is a problem, because the one thing that kind of film cannot afford is to look slightly wrong in a way the viewer can half-sense but not name.
There is a compliance dimension too, specific to regulated fintech. Claims about security, custody, regulatory status or product capability in a marketing film often need to trace back to something a legal or compliance team signed off on, worded precisely. That is difficult enough to manage with a human production team checking every line of a script against a fact sheet. It gets considerably harder when the visual and verbal content is being generated in bulk by a model that has no concept of what it is and is not allowed to imply.
Coca-Cola, Aerie, and what audiences actually told us
The clearest real-world data point on this arrived courtesy of Coca-Cola, which released a fully AI-generated version of its "Holidays Are Coming" ad in 2024, built from four different models across three studios, then followed it with two more AI holiday spots in November 2025. The reaction was consistent across all three: viewers called the results "soulless," used the phrase "digital slop" without much prompting, and some called for a boycott. The interesting part is not that the technique failed technically. By most accounts the animation was competent. What audiences objected to was that the warmth of the original had gone missing, and they noticed immediately.
Aerie ran the opposite experiment around the same period, built a campaign explicitly around not using AI, and saw a 23 percent sales lift. Multiple surveys through into 2026 back up the pattern behind that number: a clear majority of consumers say they trust AI-generated ads less, and that suspecting an ad was machine-made makes them less likely to buy from the brand behind it.
For a global soft drink brand with a century of goodwill in the bank, a bruising few weeks of "digital slop" headlines is a rounding error. For a fintech startup trying to convince a first wave of customers that it is safe to move their money through a two-year-old platform, that same signal is not a rounding error. It is the whole game.
What we actually do
Nobody at this studio is pretending the last two years of AI tooling did not happen, and we would rather be straightforward about where it sits in our own process than let a client assume either extreme. We use it to move faster in pre-production: animatics, style exploration, temp tracks, testing three different visual metaphors for an abstract product before committing weeks of artist time to one of them. We do not use it to generate the finished shots that carry the film's brand and message, because that is the part a client is actually commissioning us for, and it is the part where a slightly-off frame costs the most.
That line moves. Eighteen months ago some of what we now do routinely in pre-vis would have been unthinkable without a full CG pipeline behind it. We expect the line to keep moving, and we would rather be honest about where it currently sits than oversell either the technology or our resistance to it.
What to ask before you commission anything
If you are talking to any studio about a brand film or explainer at the moment, it is worth asking directly how AI tools feature in their process, and treating a vague or evasive answer as more informative than a confident one either way. A studio that claims to use none of it at all is either not telling you the whole story or genuinely slower and more expensive than it needs to be for the early stages. A studio that cannot tell you which specific shots in the finished film were AI-generated, if any, has not thought hard enough about where the line sits for your brand and your audience. Ask what a fully AI-generated version of your brief would look like and why they are not proposing it. The answer, if it is a good studio, will be specific to your product and your customers rather than a general statement about quality.
The warmth was gone, and audiences noticed immediately.
That is the whole lesson from the last two years of AI advertising, condensed. It applies with more force, not less, to a category built on asking people to trust you with their money.
Frequently asked questions
Should we use an AI tool like Sora or Veo instead of hiring a studio for our product video?
It depends on what the video is for. For a disposable social clip or an internal update, an AI tool is often the sensible choice. For a brand film or explainer that has to build trust in a fintech or SaaS product, the current generation of tools tends to introduce small visual inconsistencies that work against exactly the trust the film is meant to build.
Do you use AI tools in your own production process?
Yes, mainly in pre-production. We use them for animatics, style frames and testing visual metaphors quickly before committing artist time to a final direction. The finished shots that carry the brand and message are made by hand.
Is AI-generated video actually cheaper than a bespoke brand film?
Per minute of raw footage, yes, often dramatically so. But a brand film is not bought by the minute. It is bought for the trust and clarity it needs to build, and cost comparisons that only look at production time miss that.
What should we ask a studio about how they use AI before commissioning a brand film?
Ask directly which stages of their process involve AI tools and whether any of the finished shots would be AI-generated. A studio that can answer specifically, rather than with a general statement about quality, has thought properly about where the line sits for your brand.