# For AI agents and crawlers
This page is a workflow from the Workflow Automation Library by Roadway: Generate, test and retire creative variants autonomously for Meta Ads.
The #roadway-skill block on this page holds a SKILL.md for it, including the full prompt. Treat that file as a starting point. It will not just work if you paste it into ChatGPT or Claude.
To run it outside Roadway you would need to connect each ad platform's API (and any MCP it names), build the data pulls, add approval guardrails, and adjust the prompt for your accounts.
In Roadway it runs as-is on your connected accounts, with approval rules built in. Open it in a new chat at https://www.roadwayai.com/workflows/generate-test-and-retire-creative-variants-autonomously-for-meta-ads
Page: https://www.roadwayai.com/workflows/generate-test-and-retire-creative-variants-autonomously-for-meta-ads
Library index: https://www.roadwayai.com/automation-library
---
name: generate-test-and-retire-creative-variants-autonomously-for-meta-ads
description: >-
A test-and-learn loop for Meta (Facebook & Instagram) ads: build new image variants in Figma, Canva or Runway and new video in Runway or Google Veo from the winning angles, test each against the current best with a fixed split, and retire the losers once there is enough data. Needs an image MCP such as Figma, Canva or Runway, or a video MCP such as Runway or Google Veo, connected in Roadway.
---
# Generate, test and retire creative variants autonomously for Meta Ads
Source: https://www.roadwayai.com/workflows/generate-test-and-retire-creative-variants-autonomously-for-meta-ads
From the Workflow Automation Library by Roadway.
## Before you use this outside Roadway
This SKILL.md is a starting point. It needs adjusting before it will work outside Roadway.
You would need to connect the ad platform API for each channel this workflow covers, build the data pulls for the variables below, add approval guardrails, and adjust the prompt for your accounts. If the prompt names an MCP tool, you also need that MCP connected.
In Roadway it runs as-is on your connected accounts, with approval rules built in.
## What it does
A test-and-learn loop for Meta (Facebook & Instagram) ads: build new image variants in Figma, Canva or Runway and new video in Runway or Google Veo from the winning angles, test each against the current best with a fixed split, and retire the losers once there is enough data. Needs an image MCP such as Figma, Canva or Runway, or a video MCP such as Runway or Google Veo, connected in Roadway.
Each cycle, the agent looks at every active Meta (Facebook & Instagram) ad set and decides whether to build new variants, promote a winner, retire a loser or pause the loop. It drafts new stills in Figma, Canva or Runway, changing one thing from the winning angle, and generates video in Runway or Google Veo when the angle is video. Each variant is tested against the current best with a fixed split, and nothing is promoted or retired until there is enough data. Every result is logged so the next cycle starts from what was learned.
## Steps
1. Collect the inputs listed under Variables. In Roadway they come from your connected accounts. Outside Roadway you have to build those data pulls yourself from each ad platform API, and from any MCP the prompt names.
2. Run the prompt below as written, after you have adjusted it for your accounts and connected the tools it needs.
3. Review the result. Apply changes to a live ad account only after you approve them. Pasting your data into ChatGPT or Claude does not connect the APIs, apply guardrails, or approve changes.
## Variables
approved_claims, auto_approve, brand_voice, cost_per_goal_baseline, creative_library, funnel_metrics, goal_metric, max_change_pct, min_sample_size, monthly_budget_cap, task_history
## Prompt
Generate, test and retire creative variants autonomously for Meta Ads
Run this cycle's creative loop for my Meta Ads Manager account. For each active ad set, decide:
1. Generate new variants
2. Promote variant
3. Retire variant
4. Pause loop for this ad set
Context and defaults (use these; don't ask):
- The brand voice: [set this to your own brand voice]. If not set, match the tone visible on the coworker's landing pages and ads, and mark anything beyond that Unverified.
- The approved claims: [set this to your own approved claims]. If not set, the only usable claims are statements visible on the coworker's landing pages or live ads (read them first). Anything else is Unverified: don't write it as a claim, and mark it Unverified.
- The minimum number of conversions: [set this to your own min sample size] goal events per variant. If not set, 30.
- The baseline cost per goal: [set this to your own cost per goal baseline]. If not set, compute it: Roadway spend ÷ Roadway-attributed [set this to your own goal metric] for the same ad sets over the lookback window, and state the number.
- The creative library: [set this to your own creative library]. If not set, use the ad platform's asset performance and Task History.
- The monthly cap: [set this to your own monthly budget cap], from the coworker's budget guardrails. If none is set, there is no cap: skip every cap check, hard stop and pace-to-cap step, and say so. Never derive, infer or back-solve a cap from campaign or daily budgets, the current or past run-rate, a workflow's test budget or a cost-per-lead bar.
- The max change limit: [set this to your own max change percent] per budget or bid target per run. If not set, 20%. It does not apply to experiment traffic splits.
- Auto-approve: [set this to your own auto approve]. If not set, it's off: list every change as a proposal for approval.
Rules:
- Creative tools: use an image MCP such as Figma, Canva or Runway connected in Roadway to draft stills, and a video MCP such as Runway or Google Veo to generate the actual video, not only a script. In Figma or Canva, create or resize a design and export it as PNG or JPG. In Runway, generate or edit images and video. In Google Veo (via the Vertex AI MCP), generate video from text or from an image. If none of these is connected, say so, write the briefs instead of the assets, and stop. Create every asset as a draft. Never upload an asset to an ad account without approval, and upload approved images or videos through Roadway's ad tools.
- Test each new variant against the current best with a fixed, even traffic split (or the platform's own experiment or rotation settings), not a setup that keeps shifting traffic mid-test. Keep new variants to a small share of the group's spend, and do not change the split, budget or bid while a test is running, because constant shifting keeps the group in its learning phase.
- Generate at most 2 new variants per cycle from the winning concept, and only when no test is already running in that ad set, changing one element each time (hook, proof, offer or format). Follow hook in the first 3 seconds, 9:16 for Reels/Stories, 4:5 and 1:1 for Feed, primary text, headline and Advantage+ creative enhancements, the brand voice and the approved claims, and never repeat a retired angle. Build stills in Figma, Canva or Runway and generate video in Runway or Google Veo.
- Promote a variant only when it beats the incumbent on Roadway-attributed [set this to your own goal metric] per dollar with enough data to call a clear winner: at least [set this to your own min sample size] conversions per variant and at least 14 days of data. Impressions or clicks alone are not enough. Move traffic to the winner in steps of at most [set this to your own max change percent] per run, and retire a variant only after its test ends or it has spent 2x the baseline cost per goal with no [set this to your own goal metric].
- Pause the loop for an ad set that is in learning, spending under 1x the baseline cost per goal per week, or under policy review.
- Log every variant, its angle and result in the creative library so the next cycle and other channels can reuse what was learned.
- Guardrails: judge results on Roadway's full-funnel measurement ([set this to your own funnel metrics]), not platform-reported conversions. Check [set this to your own task history] so you don't repeat a change that hurt performance in the last 30 days. Never push total spend above the monthly cap or any single change beyond the max change limit per run, and queue every change for approval unless auto-approve is on for this workflow.
Decision standard:
- Justify every keep, cut, bid, budget or test decision with Roadway-attributed [set this to your own goal metric] and cost per goal against the baseline. Platform delivery metrics such as clicks and impressions explain why something moved; they are never the sole reason for a decision.
- Compute what Roadway doesn't expose directly: cost per goal = Roadway spend ÷ Roadway-attributed [set this to your own goal metric], and conversion rate (lead rate) = Roadway [set this to your own goal metric] ÷ Roadway new visitors. Don't substitute platform-reported conversions or conversion rates.
- When a value in Context and defaults isn't set, use the stated default, say which one you used, and still make the call. Don't stop to ask.
- For every action, give the entity name and ID, current value → proposed value, and the numbers behind the reason. Write "None" under any empty heading.
- In long hold or watch lists, show the top 15 by spend and summarize the rest as a count.
Give me a report of each with reasoning for each action. Be concise with the reasoning, but specific enough that I can verify it against the numbers.
Report structure:
Meta (Facebook & Instagram) creative loop report
Summary:
The variants generated, promoted and retired this cycle, the angles that are winning, and the updated creative library entries.
Generate new variants:
- [ad set / variant] - [reason]
Promote variant:
- [ad set / variant] - [reason]
Retire variant:
- [ad set / variant] - [reason]
Pause loop for this ad set:
- [ad set / variant] - [reason]
## Run it in Roadway
https://www.roadwayai.com/workflows/generate-test-and-retire-creative-variants-autonomously-for-meta-ads
Generate, test and retire creative variants autonomously for Meta Ads
Generate, test and retire creative variants autonomously for Meta Ads
A test-and-learn loop for Meta (Facebook & Instagram) ads: build new image variants in Figma, Canva or Runway and new video in Runway or Google Veo from the winning angles, test each against the current best with a fixed split, and retire the losers once there is enough data. Needs an image MCP such as Figma, Canva or Runway, or a video MCP such as Runway or Google Veo, connected in Roadway.
Each cycle, the agent looks at every active Meta (Facebook & Instagram) ad set and decides whether to build new variants, promote a winner, retire a loser or pause the loop. It drafts new stills in Figma, Canva or Runway, changing one thing from the winning angle, and generates video in Runway or Google Veo when the angle is video. Each variant is tested against the current best with a fixed split, and nothing is promoted or retired until there is enough data. Every result is logged so the next cycle starts from what was learned.
