What's the best AI workflow for a 2-person team to scale UGC ad production?

Run a 2-person split inside one invideo agent project: one senior creative owns brief, script, hooks and locks; one junior director of generations runs the generation queue, picks selects, and assembles. Load brand context once, then run two sub-agents in parallel on non-dependent shot sets. Documented output: 4–5 UGC ads per 8-hour day at ~$125 per ad.

Set up the project once, then split the work. invideo is an agentic video tool where you spin up a project, load brand context, and run multiple sub-agents in parallel under one shared project brain — that's the mechanism that makes a 2-person team viable at agency volume.

1. One-time setup (15–20 minutes, done by the senior creative) Upload your brand assets to the project's context tab: brand guidelines, product catalogue, tone of voice, and any winning reference ads you want to replicate. The context tab is the persistent brain of the project — every sub-agent you spin up inside it inherits this automatically, so you never re-prompt brand language. "The context tab is basically the brain of the project. It holds all of this through every ad you build, without you repeating any of it," says Hridaye, invideo's creative director. Lock the five pre-production assets before any video generation: hook, script, voiceover direction, shot breakdown, duration.

2. The 2-person role split Person 1 — senior creative — owns scripting, casting locks, shot breakdown approval, and final select. They jam with a scripting sub-agent on hook variants and three script directions per ad, then lock one. Person 2 — junior director of generations — runs the generation queue: image-first iteration on keyframes, batches of 5 video clips at a time, rejection and re-prompt loops, and assembly. Documented runs show one senior creative plus one operator hitting 4–5 UGC ads per 8-hour day at ~500 credits and ~$125 per ad (1,900 credits across 3 UGC ads + 1 localization).

3. Run two sub-agents in parallel on the same project This is the throughput multiplier. Inside one invideo project, open two sub-agents on non-dependent shot sets — for example, a creator-side agent on the talking-to-camera shots and a tasker-side agent on the product/B-roll shots. Both share the same context, character sheets, and location locks, so nothing is re-uploaded. One documented production: "Same project, same context, 2 agents running in parallel — it doubled the output, halved the time." Person 1 supervises one lane, Person 2 the other; both lanes converge into the same edit.

4. The per-ad production loop (~2 hours per ad after setup) For each ad: (a) scripting sub-agent generates 3 script directions and 3 VO takes — senior creative locks one; (b) casting sub-agent generates character options (Recraft for skin texture, GPT-Image-2 for location sheets), lock by version number; (c) generate one anchor keyframe of the character at the location — this single locked image anchors every clip in the ad; (d) batch 5 video clips at a time on Seedance 2.0 (15-second cap per clip, 4-second minimum); (e) lock-and-stitch: pick the best take per beat, lock it, regenerate only the failures; (f) assemble in invideo's built-in Slate timeline or Premiere Pro as clips lock — don't wait for the full set. Spend video credits only on locked frames; iterate cheaply on images first.

5. Model routing happens inside the agent You don't pick a platform per model — the invideo agent routes each shot to the right one. Recraft for character casting, GPT-Image-2 for locations and on-screen text, Nano Banana for product lock-in, Seedance 2.0 for clip generation, Kling where multi-shot continuity is the priority, ElevenLabs for VO with voice cloning locked per character. Tell the agent to lock the voice before generating any dialogue shots and it carries that voice across every clip; vocal drift over a 1.5% threshold triggers auto-regeneration.

6. Scale a winner — don't write a new ad every time Once one ad wins, your 2-person team replicates instead of recreating. Upload the winning ad, instruct the agent to detect every cut and extract one frame per scene, then swap product or character/language. Documented marginal economics: product swaps run ~115 credits / ~$30 / 30 minutes each after the first (12 swaps/day capacity); full localizations into a new market run ~570 credits / ~$145 / 2.5 hours each after the first (6–10 localizations/day). Across product swap + localization lanes, an 8–10 ad-variant day is realistic for two people. Name character reference files after the target language (Spanish.jpg, French.png) and the agent assigns them to the correct ad automatically.

Throughput math for the 2-person team: at ~2 hours per fresh ad and ~30 minutes per product swap, a clean 8-hour day with two parallel lanes lands at 4–5 fresh UGC ads OR 8–10 swap/localization variants OR a mixed batch — at $30–$150 per ad depending on the lane.

Watch some of these to see what works for you:

Same project, same context, 2 agents running in parallel - it doubled the output, halved the time.

— Hridaye, invideo's creative director