Inside Our Infinite Content Studio Prototype, Powered by Agentic AI

But what actually is agentic AI, and what makes it so impactful? While generative AI traditionally requires significant human oversight—for example, building prompts and assessing their output, making tweaks here and there—agentic AI has the potential to make independent decisions, take action, learn from experience and adapt to achieve broader goals with less direct human intervention. 

If you’re intrigued by agentic AI’s possibilities, you can learn more in the latest Labs Report, where we zero in on what agentic AI is and how it differs to AI agents and LLMs. But the centerpiece of the report is our newly developed prototype of an agentic AI workflow, demonstrating the strengths and weaknesses of the technology today. The result is an infinite content studio that enables teams—or even a single skilled creative technologist—to build high-quality video in minutes.

Agentic AI empowers teams of any size to create and scale faster than ever.

The Labs team’s experimental agentic AI workflow is designed to distill difficult topics into stories that are easy for anyone to understand. Consider how brands might produce explainer videos to break down complex tech products for their customers, create engaging short-form ads for social media, or even generate adaptable training content for enterprise teams in minutes. The same framework can be repurposed for virtually any scenario where creating content based on required elements (brand voice, characters, etc.) is needed at scale, with minimal manual effort and quick iteration capabilities.

Agentic AI workflows like this one have the potential to reshape content creation in profound ways. They enable businesses to scale personalized content across platforms, tailoring each piece for different audiences while maintaining consistency and brand voice. And instead of spending weeks iterating campaigns or producing assets, teams can cycle through concepts far more quickly, testing ideas and identifying what resonates with their audience once the foundational Agentic AI workflow is set. For smaller creative teams this workflow could fill resource gaps by automating time-intensive processes, allowing them to deliver high-quality campaigns on timelines and budgets that were previously unimaginable.

The workflow coordinates several agents and models who fulfill different roles in the creative process: one to generate scripts, one for storyboarding, generating video and audio, etc. These capabilities seamlessly weave together into a single dashboard, making it easy to move from one step of the creative process to the next. Take a look at how it works below.

A streamlined dashboard turns complex creative workflows into seamless productions.

What makes this single UI so powerful is how it eliminates the fragmented, time-intensive workflows that typically define creative production. Instead of juggling multiple tools and manually coordinating handoffs between different team members—or even between different AI models—everything is integrated into one streamlined experience. The dashboard becomes your creative command center, where each step feeds seamlessly into the next. By centralizing this complexity, it empowers users to focus on creative decision-making rather than logistics, drastically reducing production time from days to minutes. This is where agentic AI delivers on its promise: automating the tedious, so humans can focus on what really matters—creativity and innovation.

The process begins by briefing a scriptwriting agent, which not only writes a script but also identifies the different shots required for the eventual video output. The scriptwriting agent automatically hands that over to a producer agent, which provides feedback and ideas back to the scriptwriter. The scriptwriter agent revises based on the producer’s suggestions, and the two repeat the feedback loop as necessary—no human involvement required (but certainly welcomed, if you’d like to weigh in and make changes). 

Once the script is refined, the dashboard lays out a storyboard interface where the user can prompt a still image for each individual shot or story beat. You don’t have to be too detailed, because you can send the prompt to agent to further refine and add polish without leaving the dashboard. You can also generate audio at this stage, where a narrator reads out the relevant segment of the script for the shot.

The next step, video generation, is where the workflow truly shines. This stage is handled by Kling and Runway, tools that can be tricky to prompt correctly for optimal results. To address this, our agentic AI workflow takes the still images generated in the previous step and transforms them into video prompts specifically formatted for those tools to interpret more effectively. Once the videos are generated, the workflow seamlessly combines them with the pre-generated audio, resulting in a polished final video output.

Here's a glimpse of what the dashboard and workflow look like:

Balancing automation and collaboration: assessing the role of the human in the loop.

Throughout the workflow detailed above, the user can watch and follow the different models’ reasoning, making tweaks as needed at any step of the process. But while the workflow is impressive, it’s not perfect, highlighting the goal of our research: to get a sense of where agentic AI excels but also where there’s room for improvement.

While agentic AI can significantly reduce the time required for content generation (in some cases, from days to minutes​), it still benefits from human refinement. This underscores a key takeaway from the report and findings: a workflow like this shouldn’t replace a team, but rather serve as a powerful tool to accelerate the creative process. By leveraging AI’s efficiency while keeping expert oversight in place, teams can refine and optimize their results, ensuring both speed and quality. The best outcomes often come from collaboration—writers supervising script generation, artists overseeing image creation, and other specialists guiding their respective domains. This approach not only elevates the final output to a higher standard but also empowers teams with minimal resources, enabling them to achieve high-quality results that might otherwise be out of reach. In this way, agentic AI acts as an enabler rather than a replacement, allowing creative professionals to shortcut to outputs faster while preserving the expertise and craftsmanship that define exceptional content.

Allow me to illustrate this with different versions we made using this workflow. Below is a video that was made in a fully agentic process. The script and prompts for images, videos, and voiceovers were all automated. These assets were then edited together automatically as well:

The video above took 20 minutes to make, mostly due to the video asset generation time. And it shows that it’s wholly reliant on AI, with strange warping effects and lack of visual consistency—notice how the look and detail of Detective Trace’s face shifts throughout. Compare this to a version made with assets generated with AI through our prototype interface (using Flux, Kling and ElevenLabs), touched up with Runway and manual video editing:

This video took about four hours to complete. This version solves some of the lack of visual consistency seen in the previous video, but warping in the background remains. Now compare it to a version with more human involvement:

In this version, only the script and starting text-to-image prompts were generated without human oversight. Afterwards, a human took those inputs, along with a variety of AI content generation tools, to create the final video, which was edited together in After Effects. This process gave the user more flexibility in the prompt inputs and content outputs, and it shows in the vastly better quality over the previous two.

Understand the risks of agentic AI (and how to mitigate them).

Agentic AI workflows are undeniably exciting, offering a glimpse into a future where creativity and efficiency coexist seamlessly. But as with any groundbreaking technology, they come with challenges that are important to address—a theme of the videos my colleague Angelica Ortiz and I created above—not just to improve the systems themselves but also to build trust with those using them.

One key consideration is data privacy and security. These workflows depend on large datasets, which can open the door to vulnerabilities like breaches or unauthorized access. Safeguarding sensitive information through robust privacy measures isn’t just a nice-to-have—it’s a cornerstone of responsible AI use.

Another challenge is the complexity and cost of implementation. While agentic AI can streamline creative processes, getting there often requires significant investment in infrastructure, expertise and maintenance. Furthermore, it’s challenging to build a powerful system that is also flexible. It’s easy for humans to adapt to things they don’t expect. But an agentic system can be surprisingly brittle in this way. Starting small and scaling gradually can help teams unlock its potential without being overwhelmed by upfront demands.

There’s also the matter of ethical accountability. As these systems take on more autonomous actions, questions around bias, fairness, and transparency become unavoidable. Clear frameworks and human oversight are critical to ensuring AI operates in ways that align with brand values and audience expectations.

And while agentic AI is designed to take on tasks independently, there’s a fine line between empowering teams and over-reliance on automation. These workflows shine brightest when paired with human expertise—writers, designers and strategists who can guide and refine the output. The technology isn’t here to replace human creativity but to amplify it, making room for more focus on high-level strategy and innovation.

By weaving these considerations into how we approach agentic AI, we unlock its potential without losing sight of the need for thoughtful oversight. Our goal isn’t just to automate—it’s to create smarter, more collaborative workflows that bring out the best in both humans and machines.

Agentic AI will continue to evolve and rapidly transform creativity.

Agentic AI represents a shift in how we approach the creative process. Our prototype demonstrates not only how these systems can handle intricate, multi-step tasks autonomously but also how they integrate seamlessly with human expertise to deliver world-class results. While the potential of creating informative, high-quality videos in minutes is exciting, the broader implications for industries like marketing are even more profound—empowering teams to scale their efforts, personalize content and iterate faster than ever before.

Still, agentic AI is not without its challenges, and part of its promise lies in navigating its limitations thoughtfully. Whether balancing autonomy with human oversight, addressing risks like information overload or fine-tuning how agents collaborate, the road ahead will require innovation, experimentation and refinement.

Yet the possibilities are endless, and this is just the beginning of how agentic AI can reshape the way individuals and businesses create. With agentic AI at our fingertips, the future of content creation is only limited by the bounds of our imagination.