Skip to main content

Tutorial: a Product Launch, end to end

This is one continuous worked example. We'll take a single piece of work — a product launch video — from an empty system all the way to a published asset, touching every part of the Content Operations platform exactly once. By the end you'll have done the full loop.

Scenario: your team is launching a product called Aurora. The marketing manager needs a 30-second launch video produced, reviewed and published this sprint.

Roles in play: Maya (MANAGER), Dev (CREATOR), Robin (REVIEWER).

Step 0 — Be ready​

Before any AI runs, make sure you've completed Getting Started: a verified account, a funded wallet (≥ $10), a team, and your members activated with roles. Maya is a MANAGER, Dev a CREATOR, Robin a REVIEWER.

Why: AI generation is gated on verified + funded. Planning and reviewing are not — but generation in Step 4 will fail without it.

Step 1 — Create the Request​

Maya opens Manage → Requests (/requests) and clicks New Request:

  • Title: Aurora launch video
  • Type: MARKETING_CAMPAIGN
  • Priority: HIGH
  • Start / Due: this Monday → Friday
  • Story points: 8

Saving creates VID-2026-000001. This ref is now the spine everything else hangs off.

→ See Requests.

Step 2 — Decompose into layers​

From the request detail, Maya clicks Decompose and adds the layers that make up the video:

LayerKindModeAssignee
ScriptTEXTAIDev
Hero imageIMAGEAIDev
VoiceoverVOICEOVERAIDev
B-roll clipVIDEOAIDev
Background musicMUSICAIDev

Each is now a child request with its own ref, assigned to Dev. Because they have an assignee, Dev will see them in his queue; because the parent has dates, it's already drawn on the Gantt.

→ See Requests › Decompose.

Step 3 — Plan it: Gantt + Sprint​

Maya opens Manage → Planning (/planning), opens her launch project, and on the Sprints tab creates "Launch Sprint" (Mon–Fri), sets it ACTIVE, and from the backlog clicks + Add on VID-2026-000001.

She clicks 📊 Gantt view (/planning/gantt): the request already appears as a bar across Mon–Fri, showing Dev as assignee on hover. The sprint's burndown starts at 8 points.

→ See Planning, Projects & Sprints.

Step 4 — Attach a Workflow (and let AI run)​

Rather than build a pipeline by hand, Maya goes to Manage → Marketplace (/marketplace) and installs the Product Launch Campaign pack. It's cloned into the org and shows up in Workflows.

Back on VID-2026-000001, she clicks Attach Workflow → Product Launch Campaign. The pipeline looks like:

Brief ─▶ ⚡ Script ─▶ ⚡ Image ─▶ ⚡ Voiceover ─▶ ⚡ Video ─▶ ✋ Review (gate) ─▶ 📤 Publish

On attach, stage 0 runs and the ⚡ AI generate stages begin firing in sequence — script via Claude, image via DALL·E 3, voiceover via ElevenLabs, video via Runway. This is the only point credits are spent. Dev watches the stepper show "⚡ auto-generating…"; if a stage fails (e.g. a missing provider key), he uses ↻ Run now to retry.

The pipeline cascades automatically through the AI stages and then stops at Review — the approval gate.

→ See Workflows and Marketplace.

If you'd rather not use a pack, build the same stages by hand on the Workflows page — same result.

Step 5 — Track progress with standups​

While the manual touches happen, Dev posts a standup on the request: progress slider to 70%, note "AI layers generated, assembling", blockers "none". This:

  • updates the request's progress,
  • writes an immutable standup entry,
  • nudges the burndown's actual line downward,
  • notifies Maya at the milestone.

→ See Requests › Standups.

Step 6 — Approve at the gate​

Robin (REVIEWER) opens the request, reviews the assembled output, and clicks Approve at the Review gate. Approval records an APPROVED decision and auto-advances the request to the 📤 Publish stage.

Had Robin clicked Reject, the request would hold for rework instead.

→ See Workflows › Approval gates.

Step 7 — Publish​

The 📤 Publish stage runs automatically: it pushes the latest output to the configured destination and then auto-advances. As this is the last stage, the request flips to DONE at 100%.

→ See Workflows › How progression works.

Step 8 — See it in the dashboards​

Maya checks the two read views:

  • Operations (/operations) — the request now counts toward completed, turnaround and SLA are recorded, and the credits spent on the AI stages show in credits consumed. The activity feed lists every transition from create to publish.
  • AI Project Manager (/ai-pm) — with the request done, org health looks healthy; had it slipped, the timeline-risk forecast would have flagged it earlier. Maya hits ✨ AI briefing for a one-paragraph written summary of the sprint.

→ See Operations Dashboard and AI Project Manager.

What you just did​

In one loop you used every part of the platform:

  1. Request created and reffed,
  2. decomposed into AI layers and assigned,
  3. planned on the Gantt and into a sprint,
  4. drove a workflow whose AI stages auto-ran (the only credit spend),
  5. tracked it with standups,
  6. passed an approval gate,
  7. published automatically to completion,
  8. and saw it land in Operations and the AI-PM.

That's the whole content operation — and the only money that moved was the AI generation in Step 4.

Where to go next​