Introduction

The Workflow Engine is where braaand produces creatives, and where you come back to work on what it produced. That covers more ground than "make a hundred ads": a full campaign batch, a round of variations on your best-performing creative, a Score-and-refine loop that reworks the weak half of a batch, or one idea you're hand-developing. In the app it's the Workflow Engine in your brand's nav: every campaign you launch becomes a workflow there, shown as a card with a live map of its pipeline, and opening one lands you on its node canvas. One surface wears three hats: a node editor where you wire up how creatives get made, a canvas where every creative in the run is visible and pickable, and a production line that executes the graph durably, surviving closed tabs and navigation.

That combination is the point. Most tools make you choose between automating a hundred creatives and hand-developing one idea, or between shipping a batch and optimizing it. Here it's all the same act: wiring a node onto something. Wire a Score onto a whole batch, onto the three cards you selected, or onto one single creative that doesn't even exist yet. Route the losers into a Rerun, drop a Vary node on the winner to spin off new angles, or wire nothing at all and just ship what you built. The pipeline doesn't care which grain, or which stage of the creative process, you're working at.

The mental model

Four nouns, each with one job:

• A workflow (a run) is a verb: one execution of a pipeline. It owns the build's progress, its cost, its log. Find them under Workflow Engine in the brand nav.
• A board holds creatives for review: keep/discard, sections, comments, client sign-off. One board can collect results from many runs. See Boards.
• A collection organises the work: the campaign folder grouping the runs, the boards, and the creatives that belong together. See Collections.
• A share link exposes a board read-only to someone with no braaand account, the client review link. Minted per board, revocable any time.

So: workflows produce → boards hold → collections organise → sharing exposes. Every run files itself into a collection at launch and drops its results onto a board; you never have to tidy up after a run to find its output.

You don't have to be in the app to start one: a Slack channel bound to a brand can order creatives with a mention and gets them delivered back into the thread, running the same pipelines described here.

Two ways to produce

The factory makes creatives down two lanes, and they meet on the same canvas:

• On templates — the bread-and-butter lane: Templates → Brief → Build. Layouts from your shelf (or the shared library), copy written from the brief, imagery placed, everything brand-fitted. Fast, consistent, cheap per creative — this is what a plain campaign brief runs.
• From a photo — drop images into a Source node (from the library or your computer) and each becomes a creative the run starts from. That's how a photograph reaches everything the factory does: Source → Video animates it, with attempts set to a few so you can pick the take that landed.
• Straight from the image model — the AI lane: Ideas → AI Produce. Brainstorm ad concepts grounded in your Creative Context Library, then generate one finished ad per concept with no template at all, grounded in the brand two ways: a rendered brand board (the real logo, exact colors, the brand's own letterforms) rides along as the reference image, and your written guidelines ride in the prompt. Chain a second AI Produce onto the results to refine the takes, put a Hold between the brainstorm and the spend so a person approves the concepts first — and when a generated ad earns a future, Make editable rebuilds it into a real document whose layers are the brand's actual logo, editable text in the brand's faces, and a real button.

Score, Route, Boards and Slack sit downstream of both. A generated ad is an ordinary creative the moment it exists, so the two lanes mix freely in one graph — the node reference covers every block in both.

Three ways to work

Everything on the canvas can be wired at three grains, and they're not three features. They're three places to start a wire:

1. A whole batch: every lane of creatives has an outlet on its right edge meaning "everything this node produces". Wire it onward and the full batch flows.
2. A selection: check a few cards and a node's Run narrows to exactly those.
3. One creative: a single card (or a promise of one, see the node editor) wires out on its own, takes its own path, and genuinely leaves the batch. Whatever the batch feeds next works on the rest.

Runs are durable

Press Run pipeline and the run belongs to the server, not your tab. Navigate away, close the laptop, it keeps building, shows up as a live pill next to the command palette orb wherever you are, and pings you when it's done. Reopen its canvas any time to watch each creative move through its stages (writing copy, finding an image, fitting layout), or come back a week later and wire something new onto the finished batch.

A run's full accounting, what it built, what each stage cost in time and credits, what the judge said, is one click away: Copy run log in the canvas sidebar produces a shareable text log you can paste to a colleague or an AI.

Where to go next

• Campaigns: the quick path, hand braaand a brief, review the batch on a board.
• Workflow builder: promises, wiring, and how the canvas works.
• Nodes: every building block, one by one.
• Examples: worked pipelines to copy from.
• Boards and Collections: where results land, and how they're organised.
• Recipes: reusable pipelines you run by name.