Use case · Creative production
Turn product photos into UGC-style images and short video ads in Chat. Every file is saved with its prompt, and the brand team approves it.
An illustrative creative production use case for Agent Serve, the AI workspace where teams build, deploy, and manage AI agents: image and video generation in Chat, media models pinned per chat, cost confirmations, Library provenance and a brand review workflow.
At a glance
The problem it removes
Each new ad needs a shoot or a designer’s afternoon. Teams that try AI image and video tools end up with files in separate accounts, no record of the prompt or the cost, and no clear point where the brand team signs off.
Step by step
A designer attaches the product photo and describes the scene, the person and the format.
The designer pins an image or video model for the chat, kept apart from the model that reasons.
Drafts wait for a brand reviewer before anyone uses them in an ad.
How it fits together
Designers brief in Chat with the product photo as a reference and get images or video back.
Media models come from a live catalogue, such as Veo 3.1, Kling v3.0, Seedance 2.5, Nano Banana 2.1 or FLUX.2.
Every generated file lands in the Library with the prompt, the model and the cost.
A review workflow pauses at a Human in the loop step until a brand reviewer approves.
Access policies set allowed models, default media models and the cost confirmation threshold.
How it runs
Agent Serve generates the drafts and keeps their history, and a brand reviewer decides what goes into an ad.
The product photo goes in as a reference, for image-to-image or as the first frame of a video.
Under + → Media models, the designer pins a video model for this chat.
A generation estimated above the workspace’s threshold asks before running, then shows progress with a cancel button.
The video lands in Library under Generated/Videos with its prompt, model and cost.
What to measure
Example figures for a team producing ads for a catalogue of a few hundred products.
4×
Variants per product
More angles to test without another shoot.
−50%
Time to a first draft
From product photo to a reviewable ad.
100%
Files with provenance
Prompt, model and cost saved with each file.
100%
Assets brand-reviewed
Nothing is used before approval.
Illustrative example. Figures show how the workflow is measured; they are not customer results.
In depth
Agent Serve puts the brief, the models, the review step and every generated file in one workspace.
Agent Serve Chat generates images, video, speech and music from a written brief and reference images.
An Agent Serve workflow drafts variants for each new product, then pauses at a Human in the loop step for the brand reviewer.
The reviewer sees the draft and approves or declines it, and the decision stays with the Agent Serve run.
Examples
Three creative teams producing ads on Agent Serve, each with a brand reviewer approving before anything is used.
An Agent Serve chat turns a flat-lay photo into images in the style of customer posts, saved to the Library for review.
An Agent Serve chat uses the product photo as the first frame of an 8-second video for paid social.
An Agent Serve workflow drafts lifestyle images for each new product and waits for the brand reviewer.
Examples show what teams can build in Agent Serve. They describe patterns, not named customer deployments.
Before and after
| Work | Before | With Agent Serve |
|---|---|---|
| New angles | BeforeA reshoot or a designer’s afternoon per variant. | With Agent ServeDrafts generated in Chat from the product photo. |
| Where files live | BeforeDownloads from separate AI tools on personal accounts. | With Agent ServeThe Library, with prompt, model and cost on every file. |
| Spend on generation | BeforeFound on the card statement at month end. | With Agent ServeGenerations above a threshold ask before they run. |
| Brand sign-off | BeforeA thumbs-up in a chat thread, if anyone asks. | With Agent ServeA review step that records who approved each asset. |
Governance
Agent Serve records how each file was made and lets admins decide which media models a team may use.
Security and governanceAgent Serve access policies can turn media generation on or off and set default image, video, speech and music models.
Admins set the threshold above which an Agent Serve generation asks for confirmation before it runs.
Relays never send email or book meetings without an approval, under any autonomy setting. An approval covers the exact payload the reviewer saw and expires if nobody decides.
Proposals, approvals, declines and executed actions are written to the workspace audit log, which admins can export.
Admin, write and read roles per workspace, plus permission groups that limit which models, blocks and integrations a team can use.
Use your organisation’s own model provider keys at organisation or workspace level.
Every workflow run records each block’s inputs, outputs and errors, so a reviewer can trace how a result was reached.
Further reading
The mirror selfie, the cafe-table flat lay, the street shot mid-stride. This playbook shows how to recreate the look of the shots your audience saves, with your own garments in them and a person approving every image.
Fashion · 6 min readEvery new product needs a few seconds of motion for Reels and TikTok. This architecture turns one approved still into vertical and square drafts with a voiceover script, and nothing goes out until a person approves it.
D2C brands · 5 min readAgencies juggling a dozen brands end up with a dozen logins to separate AI tools and no idea what each client cost. Here is one production line: brief in, reviewed creative out, spend tracked per client.
Creative & content agencies · 5 min readQuestions
Agent Serve lists media models from a live catalogue, such as Veo 3.1, Kling v3.0, Seedance 2.5, Nano Banana 2.1, Seedream 5.0, FLUX.2 or GPT Image 2. The catalogue changes, and admins can limit which models a team may use.
Not in one click. An Agent Serve workflow runs per product, triggered by each new product or run over a list, and every draft still waits for brand review.
A generation estimated above the cost confirmation threshold asks before running. Admins set the threshold in Agent Serve access policies, and each file records its cost.
No. The figures are illustrative examples of what a creative team might measure with Agent Serve, not customer results.
Last updated
Enterprise access
Agent Serve is available to teams by invitation. Tell us what you want to build and we will set up your workspace with you.