AI Marketing Needs Human Taste, Not More Tools

There is a question more marketers are starting to ask whenever they see a polished launch video, a slick campaign asset or a brand film with cinematic confidence:

“Was that made with AI?”

It is a fair question. AI has moved from novelty to normal in marketing. HubSpot’s 2026 State of Marketing report says 80% of marketers use AI for content creation, and 75% use it for media production.

So yes, the IntentOne launch video was made with AI.

Watch the video, read the breakdown.

This is the video used to launch IntentOne. Watch for the human choices: the mood, the restraint, the refusal to use obvious images, and the way the video launches a category, a product and a narrative.

But that is the least interesting part of the story.

The better question is this: what did the humans do?

Because the video was not the result of someone typing a prompt and waiting for a miracle. It was not a clever shortcut. It was not an agency-style production with a few AI effects added afterwards.

It was something more revealing for the future of AI Marketing.

Agents ran much of the production. Humans directed the truth, the taste and the judgement.

That distinction matters. Because as AI makes the cost of trying things collapse, the value of knowing what is good goes up.

The real brief was not “make a video”

IntentOne needed more than a launch asset.

The video had to carry a refreshed brand, introduce a product and signal a new category: Opportunity Intelligence.

More than that, it had to give IntentOne a distinctive place inside that category. It had to do this quickly, clearly and memorably, without feeling like a conventional product video.

That is a lot to ask of a short video.

A standard product explainer would not work. A feature walkthrough would have been too small. A lead-generation video would have pulled attention to the wrong place.

The job was not simply to explain IntentOne.

The job was to make people feel that something new had arrived.

That is where the strategy started. Not with a tool. Not with a storyboard. Not with a production partner.

It started with a diagnosis.

Enterprise AI brands are collapsing into one another. The same words. The same claims. The same glassy visuals. The same “AI-powered” promise. The same confident, abstract language that sounds impressive until you realise everyone else is saying it too.

For marketers, this is the uncomfortable truth: AI makes it easier to create. It also makes it easier to sound like everyone else.

That is why the launch video needed to turn heads. It needed to feel different from the usual stream of AI tech content. It needed to carry a mood, a point of view and a sense of intent.

So the video had to avoid the category’s own clichés.

It had to launch an AI-native product without sounding like every other AI-native product.

That is a harder creative problem than it first appears.

The agency alternative was not compelling

The obvious route would have been to brief agencies.

But the options on the table felt familiar. Many were offering saturated stop-motion style concepts at prices that did not match the ambition or the moment. There is nothing wrong with using an agency. In many cases, they bring valuable craft, perspective and production power.

But this project needed something else. It needed speed, control and intimacy with the message.

As an AI-native company, the more natural question was: could we make this ourselves using AI well?

Not casually. Not by throwing a prompt into a tool and hoping for magic. But by combining human direction with AI-enabled production.

That distinction matters.

AI Marketing is not just about using tools. It is about building the conditions in which those tools can produce work that is on-message, on-brand and worth someone’s attention.

Good strategy closes doors

A lot of marketing strategies are polite lists of things a team might do.

This one worked because it ruled things out.

No agency.

No generic text-to-video.

No old phrases that the brand had decided to stop using.

No obvious picture for the obvious line.

Those exclusions mattered.

Text-to-video was tested, but it produced work that felt too generic and too far from the brand. So it was cut. That decision hurt, because text-to-video would have been easier. But strategy only becomes real when it costs you something.

The same principle shaped the language. Before a word of the script was written, the team locked a source of truth: one brand truth, one positioning line, one product sentence and a list of retired phrases that were not allowed back into the work.

It started with three foundations

The launch video began with three critical components: a prompt, a design system and a knowledge base.

A prompt

The prompt gave the work its initial direction.

A design system

The design system gave the work its boundaries.

A knowledge base

The knowledge base gave the work its truth.

The prompt gave the work its initial direction. The design system gave it boundaries. The knowledge base gave it truth.

That third part is often overlooked. A large language model can produce fluent content very easily. Producing content that is strategically correct is another matter. For that, it needs access to the right context: positioning, messaging, product language, customer pain points, category framing and brand voice.

The knowledge base was not a folder

That last part is important.

Most knowledge bases remember what a company has said. This one also remembered what the company had decided to stop saying.

That is how the script avoided drifting into the broad, shiny, generic register of AI marketing.

The system was not just fed context. It was given a canon.

For the IntentOne video, the knowledge base became an enforced message map.

It drew from internal sources, but the important part was not the volume of information. It was the discipline around it.

The system knew the brand truth. It knew the positioning. It knew the product sentence. It knew the banned language. And it helped create a script that stayed on-message before a human editor tightened it for time, rhythm and emotional force.

That is a useful lesson for any marketer using AI.

A weak knowledge base makes fluent content.

A strong knowledge base makes faithful content.

An enforced canon makes content that can survive pressure.

This was agentic production. Humans directed.

The creative direction came from people. Myself, and Jerry Marcel Lieveld, our VP of Design.

There was a clear sense of the cinematic quality required. The video needed to feel serious but not cold. Confident but not bombastic. High-stakes but not self-important.

That nuance matters because AI tools do not automatically understand the emotional context of a launch. They do not know what it feels like to introduce a new category while refreshing a brand and bringing a product into market at the same time.

Now here is the part that is easy to miss.

Nothing was shot.

There was no film crew capturing footage and then using AI to enhance it. The visual layer was synthetic. Stills and motion plates were created through an agentic pipeline, with software agents helping assemble prompts, generate frames, animate them, grade them and check them.

Humans judged the work at each gate.

AI agents can generate options. They can accelerate production. They can help a small team move with the force of a larger one.

But they cannot replace taste.

And taste was the real differentiator.

That matters because it changes the creative model. AI was not brought in at the edge of production. It sat inside the production engine.

But the humans still made the crucial calls.

They decided what the video needed to feel like. They rejected work that looked easy but wrong. They kept the visual language aligned to the brand. They decided when a content filter refusal was not just a blocker, but a better creative path.

This is where AI Marketing starts to look less like “using tools” and more like conducting a system.

The design system became executable

Traditional brand guidelines often sit in a PDF.

They tell people what the brand should look like. Then, under deadline pressure, people interpret them, bend them or ignore them.

This project needed something stronger.

This is where Jerry’s work became central. The design system was not just a visual reference point; it became part of the production logic.

The design system acted less like a style guide and more like a generative runtime. It did not simply check visuals after creation. It helped generate visuals within the rules from the start.

Each scene prompt came from one brand-prompt source of truth. The colour grade and constraints stayed consistent. Only the scene changed.

The visual process followed four stages.

That sequence mattered.

The still locked the brand-correct frame before motion entered the process. Text-to-video, by contrast, kept pulling the work towards generic AI imagery. Image-to-video inherited the discipline already present in the frame.

The design system also handled cultural and visual specificity. Casting was routed by city demographics rather than generic global diversity. Brand accents appeared as surfaces in the world: a painted shopfront, a lattice screen, a curtain, a real-feeling detail. Not a colour wash laid over everything.

That is the difference between branding as decoration and branding as architecture.

The video refused the obvious image

One of the strongest creative decisions was also the simplest to describe.

Never illustrate the line with the obvious picture.

That rule gave the film its edge.

When the voiceover suggested that an opportunity had gone, the film did not show a lost-sale graphic. It showed empty chairs after a meeting. A closed curtain. A place someone had just left.

The image sat at an angle to the words.

That angle matters because it invites the viewer to complete the meaning. It gives the audience a small act of discovery. It also avoids the on-the-nose visual language that AI tools tend to produce by default.

This is one of the clearest lessons for marketers.

AI often gives you the most literal version of an idea. The work is to pull away from it.

Literal is fast.

Obvious is easy.

Memorable usually sits one step to the side.

Music came before motion

The music was generated with Suno, an AI music generator that creates original music from descriptive prompts.

But the music prompts were not casual. They came from the same reasoning and creative intent behind the rest of the film. Suno rendered the score. It did not invent the intent.

An early track created the right atmosphere, but it did not yet match the story. So the music was chopped, shaped and synced with the script before the graphics were created.

That workflow inversion was vital.

Instead of making visuals first and dropping music underneath, the music became the spine. It gave the film rhythm. It shaped the emotional rise and fall. It helped define the cut before the final motion plates existed.

For marketers, the lesson is simple: AI assets are not finished outputs. They are material.

• Cut them.
• Test them.
• Reject them.
• Rebuild them.
• Make them serve the idea.

Voice was trust, not narration

The voiceover used ElevenLabs, which provides AI voice generation and text-to-speech tools.

But the voice was not chosen from a menu and dropped into the edit.

It was constructed.

Multiple candidate voices were auditioned and scored against a defined target: trusting, independent and lightly melancholic. The team then worked from a voice doctrine: length, breath, range, mastering and signature.

The final read was built phrase by phrase. Pauses were placed by hand. Silences were treated as part of the performance. The audio was mastered. The pitch swing was checked against a target to make sure the read carried enough emotional range.

That sounds detailed because it was.

And it should be.

In AI Marketing, voice can make or break trust in seconds. The wrong voice makes the work feel synthetic, even when the words are right. The right voice gives the brand room to breathe.

A voiceover is not just a layer of sound.

It is a signal of character.

Assembly is where meaning appeared

Jerry cut the final video in DaVinci Resolve, bringing the generated motion plates, product capture, brand graphics, Suno score and voice master into one coherent story.

This was the craft stage.

Generated motion plates, product capture, brand graphics, the Suno score and the voice master had to become one coherent story.

Some overlays helped tell the story. Others hid small imperfections in the generated footage.

That is not a confession. That is editing.

Every production has seams. The craft lies in knowing where they are, what matters, and how to guide the viewer’s attention towards the idea rather than the machinery.

This is also where the film became more than a set of AI-generated parts.

AI made assets.

Editing made rhythm.

Judgement made meaning.

What was genuinely advanced?

It would be easy to overclaim.

The tools themselves are not rare. Many marketers can access AI image tools, video models, Suno, ElevenLabs and DaVinci Resolve. The field has moved quickly.

High-resolution visuals, audio sync and cinematic polish are no longer enough to make a piece of work exceptional.

So what was genuinely hard to copy?

Not the tool stack.

The taste layer.

The strongest parts of the process were the systematic refusal to be literal, the culturally specific design runtime and the governance that kept the production on-brand at every gate.

That combination is still uncommon.

Many teams can generate video. Fewer can govern a synthetic production with a locked truth, a banned-phrase canon, prompt discipline, visual rules, voice scoring, content-filter pre-checks and human judgement built into every stage.

The point is not that AI made the work effortless.

It made each attempt cheaper.

That created a new pressure: when trying becomes cheap, choosing becomes the job.

What did it cost?

The unglamorous answer is iteration.

Many more visuals were generated than made the cut. Stills were re-rolled when the colour lacked the right intensity. Motion clips were rejected. Some ideas failed content filters and had to be restaged.

One rejected close-up handshake became empty chairs after a meeting.

That was not just a workaround. It was better. More observed. Less staged. More in keeping with the rule against obvious illustration.

This is a very modern creative truth.

AI reduces the cost of the attempt. It does not remove the cost of taste.

Someone still has to judge. Someone still has to say no. Someone still has to know when the “failed” version has led to a better idea.

What marketers should take from it

The IntentOne video offers a useful model for AI Marketing teams.

• Start by locking the truth before you generate anything. Do not let the model decide your positioning for you.
• Make your knowledge base remember what you no longer say. Old language has a habit of finding its way back into new work.
• Anchor motion to a strong still. A brand-correct frame gives video generation a better starting point than raw text.
• Treat AI outputs as raw material. The first version is rarely the answer.
• Fight the literal. AI will often give you the obvious picture. Your job is to find the truer one.
• Make the design system executable. A brand PDF advises. A generative system constrains.
• Plan for judgement. The scarce resource is no longer the attempt. It is attention.

The honest tension

The film worked. But the best learning comes from looking at the tension too.

On-screen text sometimes repeated the voiceover exactly. That helped reinforce the message, but it also worked against the non-literal imagery. In a few moments, trusting the picture more may have made the film even stronger.

There was also a representation watch-point. The casting was global and specific, but some scenes risked coding one part of the world as crowds and another as individuals. That kind of pattern can appear quietly, even inside careful work. It deserves deliberate checking next time.

This honesty matters.

A case study that only flatters itself becomes a brochure. A useful story shows the method, the wins and the next questions.

We had great fun making it, but it was not effortless. It was full of judgement calls, failed attempts, unexpected fixes and moments where the better idea appeared only after the easier one broke.

The future of AI Marketing is not tool-first

The IntentOne launch film was made with AI. But it was governed by humans at every layer.

• A locked truth.
• A council of models used to test concepts.
• A voice tuned like an instrument.
• A design system that generated rather than advised.
• A refusal to say the obvious thing with the obvious picture.
• An edit that turned assets into meaning.

That is the lesson for marketers.

The future of AI Marketing will not belong to the teams with the longest tool list. It will belong to teams that can combine clear strategy, strong taste and executable systems.

AI can make every attempt feel almost free.

That is exactly why judgement becomes priceless.

So the next time a launch film, campaign or brand asset makes you ask, “Was that made with AI?”, ask the better question.

What truth did the humans make unforgettable?