A workflow chains platform actions — document audit, structural analysis, review, fixes, merge, publication — with AI agents and a human approval at every binding step.


A real flow from the platform: six nodes, a bounded loop and a human approval before anything is written to the client’s system.
The site’s documents are read and structured.
Run when the audit succeeds and the structure is a lattice tower.
Control rules: if the result is compliant and the score high enough, the site moves on to approval.
On a blocking anomaly, fix then review again, at most twice.
The flow waits until an engineer has reviewed and approved the branch.
The site is pushed to the client’s document management system.
Drag nodes, connect them, set each action and its parameters.
Run on a selection of sites or on an event; each edge carries its condition (e.g. “status = success AND structure type = lattice tower”).
Review → fix → merge → review, with a maximum number of passes: beyond it, the run stops as “blocked”.
Wait for every predecessor or resume on the first one.
Each flow is versioned and can be enabled; you choose who may start it.
Every run is tracked site by site: progress, status, history.
Each node is a platform action, with its parameters and permissions.
The flow waits on a task: an engineer reviews the diff and decides.
The flow assigns a task to an AI agent and waits for its outcome.
Discover AI agents →



A workshop with your teams, then two days of building on your real data.