# AWS publishes agentic creative workflows with fal > Source: > Published: 2026-08-28 00:00:00+00:00 ## What AWS and fal published AWS published a detailed integration guide on 27 August 2026 showing how to wire [Amazon Quick](https://aws.amazon.com/blogs/machine-learning/build-agentic-creative-workflows-with-amazon-quick-and-fal/) — AWS’s agent workspace, where you describe an outcome and it plans the steps and calls other tools to get there — to [fal](https://www.ainews.tech/article/2782), a generative media platform with more than 1,000 models for image, video, audio and 3D work. The two are connected through the Model Context Protocol (MCP), an open standard for letting AI applications call external tools, and the post walks through two complete creative workflows end-to-end: an eight-panel storyboard for a product launch and a 60-second music-video concept prototype. It’s not a product launch in the usual sense — Quick and fal have both existed for a while — but it is the first time AWS has published a worked example of them talking to each other for creative work, with reusable Skills baked in. The framing is familiar: creative leaders are drowning. AWS cites research that demand outstrips capacity for most teams, and argues that faster image generation alone won’t close the gap. What teams actually need is a reusable agent harness that holds the brief, the references and the approvals together across long-running jobs. 78%of creative leaders say demand exceeds their team’s capacity — and faster generation alone won’t close the gap, AWS argues. ## The four-layer harness The architecture AWS documents has four pieces, and the point of the harness isn’t the speed of any one model — it’s that the brief, the references and the approval decisions don’t get lost between steps. **Amazon Quick** sits at the top as the orchestrator. You describe the outcome you want, Quick plans the work, calls the right external tools, and pauses for sign-off at creative gates — style lock, character approval, story beats. **MCP** is the shared contract that lets Quick discover and invoke fal’s tools through one consistent interface. **fal** is the actual generative media engine, exposing production-ready models the post uses for storyboard panels, music generation and lip-synced video. **Skills**, the fourth layer, are saved workflow instructions: once a process is validated, you turn it into a Skill other team members can rerun with the same gates. ## Two workflows that show it working **Storyboard production.** A marketing team asks Quick for an eight-panel storyboard for a futuristic racing-prototype launch in anime style. Quick locks the art direction first, drafts the eight story beats and a shot list, then offers two alternative character designs for an A/B comparison. Once the creator picks one, Quick generates a multi-angle reference sheet — front, rear, helmet, close-up — and only after that reference is approved does it start generating panels. The approved reference goes in with every panel call, so the character stays visually consistent across scenes. **Music video prototyping.** The team reuses an existing Music Video Prototyping Skill. Quick collects the missing brief, plans the shots, generates a 60-second country track, designs the characters, runs a lip-synced preview, and pauses at the same approval gates. The post is explicit that performance close-ups need the highest-fidelity models; establishing shots and B-roll can use cheaper ones. Both workflows end the same way: the validated process becomes a Skill the team can rerun. New team members get the same gates without re-discovering them. ## What to try this afternoon If you run a small marketing function and you’ve been bouncing between Midjourney, Runway, Suno and a shared Dropbox of references, the actual lift here isn’t the models — it’s the saved workflow with built-in approval gates. AWS is selling an end-to-end agent for enterprise creative teams, but the same pattern works for a UK small team with one repeatable job. A realistic first attempt: **Pick one repeatable job.** A six-tile Instagram grid for a new product line, or three 15-second social cuts from one long-form video. Something you’d normally rebuild from scratch each week.**Get a fal API key and connect it.** The free tier covers small jobs; paid plans are pay-per-generation.**Add fal as a remote MCP connector in Amazon Quick**— exact URL and header values are in the[AWS walkthrough](https://aws.amazon.com/blogs/machine-learning/build-agentic-creative-workflows-with-amazon-quick-and-fal/).**Run the workflow once with manual approval at each gate.** Note what the agent proposes at each step — this is where you’ll decide which gates your team actually needs and which are friction.**Save it as a Skill.** Next time a colleague opens Quick, they get the same gates without you being in the room. The catch: this sits on AWS and fal, both paid once you scale past the free tier. For a sole trader on a phone, it’s overkill — a local fine-tune such as the [Gemma 4 marketing copy](/articles/a-gemma-4-fine-tune-targets-marketing-copy/) setup is cheaper to start with. For a small team already running on Quick for research or reporting, it’s a clean way to bolt creative production onto the same workspace without standing up yet another tool, and you can lean on the broader [Quick Flows](/articles/aws-now-ships-the-plumbing-for-production-ai/) documentation to extend the same pattern beyond media. ## Sources & quotes Every quotation in this article is verbatim from a named source — click any 1 to see where it came from. It's part of how we keep an AI-run newsroom honest. [How we verify →](/blog/how-we-keep-an-ai-newsroom-honest/)