Slack concept studio: three AI image options from any brief

Build an agent-based workflow that turns Slack briefs into three AI-generated concept images and archives the team's pick into a dated folder in Google Drive.

Trigger: an incoming Slack webhook. The workflow should fire whenever a teammate either uses a /concept slash command or @-mentions the workflow in a marketing or design channel. The brief is the text the user types after the command, or the body of the mention message. Capture the channel id, the thread parent message ts, and the user id of whoever invoked it.

Step 1, generate three distinct variations. Call generate-image three times against the same brief, but ask the model for meaningfully different directions, not minor tweaks. Vary at least style (photoreal, illustrated, minimalist line work, 3D render, collage, etc.), composition (close-up vs wide, centered vs rule-of-thirds, isometric, etc.), and color palette. For each generation, also produce a short one-line note describing the direction so the team understands why each option looks the way it does. If the brief is empty or too vague to act on, reply in the same thread asking the user for one extra sentence of context and stop. Do not guess.

Step 2, reply in the same Slack thread with all three options. Use the Slack Upload a File operation three times, once per variant, each as a reply on the thread_ts of the triggering message. Each upload should include a caption like 'Option 1 — bold editorial photoreal' with the variant number and the one-line direction note. After the three files are posted, send one more Slack Send a Message reply in the thread asking the team to react with 1, 2, or 3 (number emoji) or 👍 on their favorite, and to react on the file post for the option they want to keep.

Step 3, watch for a reaction. Poll the three uploaded variant messages using the Slack Get Reactions operation. Treat the workflow as having a winner when any of them collects a 'one', 'two', 'three', '1️⃣', '2️⃣', '3️⃣', 'thumbsup', or '+1' reaction. If multiple variants get reactions, prefer the one with the most reactions; break ties by most recent. If no reaction arrives within a reasonable window (default 24 hours), leave the three options in the thread, post nothing further, and stop. Do not nag.

Step 4, archive the chosen image to Google Drive. First, make sure a top-level Brand Library folder exists: use Google Drive List Files with a q like "name = 'Brand Library' and mimeType = 'application/vnd.google-apps.folder' and trashed = false" to find it, and if it does not exist, create it with Google Drive Create Folder. Then inside Brand Library, create a dated subfolder named like '2026-05-31 launch hero', where the date is today and the slug is a short kebab-case version of the brief. Again, List Files first under the Brand Library parent, then Create Folder if missing. Finally, use Google Drive Upload File (Multipart) to put the chosen image into that subfolder, with a filename like '<brief-slug>-variant-<n>.png' and the right image mime type.

Step 5, confirm in Slack. Once the upload succeeds, send one final Slack message in the same thread with a short confirmation and a link to the new Google Drive folder, so the team can find the archived asset without digging through Drive themselves.

Integrations to wire up: Slack (for receiving the slash command or mention, uploading the three variant files in the thread, reading reactions, and posting confirmations), generate-image (for the three image variations, built into the platform), and Google Drive (for finding or creating the Brand Library and dated subfolder, and uploading the winning image).

Operating principles: keep the conversation in the original Slack thread, never DM the user. Never post the variants to a different channel. If any single step fails (image generation, upload, Drive write), post a short, plain message in the thread explaining what failed and stop, instead of retrying silently.