I Built an AI Agent Team: From Finding Trends to Producing a Video

Hi, I am Guozhen.

Many readers are interested in making videos. I often receive messages asking how to build an AI agent team that can run the full video-production pipeline.

In this article, I will share my own workflow.

1. Result demo

The core video-production process is roughly this:

First find trends, then decide whether a topic is worth making. Next write the script, design the storyboard, generate visuals, produce the video, and finally add subtitles.

If every step requires copying text, switching tools, and pasting results back and forth, the process is still tiring.

The core workflow can be handled by building an AI agent team. Today I will introduce one platform I have used: OmniWork.

It requires no coding and the workflow feels smooth. Let us look at the final result first.

I asked OmniWork to find recent AI topics suitable for content creation, then choose one direction and continue turning it into a video.

What it gave me was not just a suggestion. It kept pushing the process forward and finally produced a video.

It can generate a video smoothly through an agent team, and it can also produce a competitor-capability radar chart from one sentence:

After running the whole process, my biggest feeling was this: it is not merely helping me "think of an idea"; it is completing the work.

Below, I will walk through how each agent member in the team is built.

2. Build the video-generation team

Step 1. Let the first member of the content team start: the trend-finding agent

For real content work, the first step is not writing. The first step is judging what is worth doing right now.

I first called the expert called Trending-Content-Monitor:

The task I gave it was roughly:

Please monitor recent hot topics in the AI large-model space and filter out three topics suitable for short videos.
For each topic, provide the trend source, why it is worth making, and whether it can be turned into a video.

As shown below, the large model used here was Kimi K2.6:

The expert started working and called its built-in web-search skill:

It organized content directions from a trend perspective instead of directly giving me generic copy.

It then output three topics:

For many people running a public content account, the missing piece is not an AI writing tool. The missing piece is someone who helps judge the direction.

Step 2. After choosing the topic, call the Film-Production-Director expert

Once the topic is selected, the next handoff is video generation. I directly called the Film-Production-Director expert without jumping to another page.

The task was:

Based on topic one, "Google I/O 2026 AI product family overview," create a video.
Requirements: complete script, storyboard design, character and scene settings, video generation, and final video output.
The process can move forward directly without asking me to confirm each step.

Then it generated the storyboard. For example, what each shot should show, where the character stands, how the camera moves, and what the visual atmosphere should be:

It then generated a second storyboard video and automatically connected it with the first one:

This is important for video content. A script alone is not enough; the visuals also need to stay consistent.

It continued from script to storyboard, scene design, and final video. The screenshot of the final video is below:

The video is not instantly at cinema quality, but from the perspective of a self-media creator, it is already usable, especially considering that most people are not video-production experts.

Previously, making a similar video required visual creation, editing, and voice work. OmniWork compresses these steps into one expert workflow.

Step 3. Send it to the Professional Video Editor expert for subtitles

Next I handed the result to the Professional Video Editor expert. Again, there was no page switching; it directly called the expert to add subtitles:

After a short wait, subtitles were added:

The full video was shown at the beginning of the original article. In addition, it can convert the video to 9:16 and perform automatic editing.

3. Competitor-analysis agent

Besides video production, I also tested a scenario closer to my daily writing workflow: analyzing AI Agent products such as OpenClaw, Hermes, OpenHuman, and Claude Code.

To help me write product-analysis articles, I called OmniWork's content-analysis expert:

This scenario is very useful for content creators. When writing a product-analysis article, the most time-consuming part is not the writing itself. It is research, information organization, horizontal comparison, and forming a clear point of view.

Here is the work process:

OmniWork did not tell me "how competitor analysis should be done." It directly organized the tables, charts, and conclusions:

Comprehensive delivery-capability chart:

And a matrix chart:

Final thoughts

After this test, I feel OmniWork is not only useful for video. It can also extend into product research, content planning, and business analysis.

It does not just give advice. It outputs tables, charts, and videos, which is closer to an "all-in-one AI expert team" for content work.

Its core value is not single-point material generation. Through Experts, Skills, Autowork, and memory, it can keep pushing a creative goal toward delivery.

For space reasons, I only showed part of the workflow here. You can try it yourself:

www.omniwork.ai

OmniWork is still in beta. If you are interested, it is worth testing.

This English edition preserves the screenshots and workflow order from the original Chinese article.

Final verdict

The interesting part is the agent-team chain: trend discovery, material organization, and video output belong in one workflow. That is the direction content automation needs to move in.