Loading
The Command Centre is the operator console of the BlueGreen governed commerce operating system. One screen shows what your AI agents did, what needs a human decision, and what changed for every tenant — commerce command orchestration, shelf audits and stack operations each in their own lane, with every high-risk action queued for approval and evidence attached.
A command centre for ecommerce is the operator console where a commerce team monitors automated work, approves high-risk actions and audits what changed. BlueGreen\u2019s Command Centre presents three lanes — Commerce Command, ShelfCheck and Commerce Stack — so one person can supervise governed automation across every store and tenant from a single screen.
The Command Centre is organised into three lanes, and each lane is a governed window into one part of your commerce operation. The commerce_command lane is where commerce command orchestration happens: catalogue synchronisation, product content workflows, pricing checks and support drafting all report their status here. The shelf_check lane carries ShelfCheck's continuous product content audits, from the baseline scan to ongoing listing-rule checks. The commerce_stack lane tracks the consolidated commerce-operations jobs that run on the platform's shared data spine — the work that used to be five separate apps.
Lanes exist so operators never context-switch blindly. An agency account manager opens one tenant, sees that the Commerce Command lane has three approvals waiting and the ShelfCheck lane has run clean, then moves to the next tenant with a single keystroke. Nothing is buried in a per-app dashboard, because there are no per-app dashboards — one screen, three lanes, every tenant.
| Lane | What it runs | What the operator does |
|---|---|---|
| Commerce Command | Commerce command orchestration: catalogue sync, product content, pricing checks, support drafting | Approves high-risk changes, monitors workflow health, reads daily operating summaries |
| ShelfCheck | Continuous product content and shelf audits against channel listing rules | Reviews audit reports, approves drafted corrections before publication |
| Commerce Stack | Consolidated commerce-operations tooling running on one data spine | Sees which jobs replaced which apps, tracks sync and integration health |
Governed automation only works if the human gate is honest, so the approval queue is the most important surface in the Command Centre. When an agent proposes a high-risk action — a price change above your threshold, a bulk edit past your SKU limit, a refund, anything touching customer data — the action queues with its evidence attached: the diff of what would change, the catalogue or order data the AI acted on, and the rule that triggered the gate.
Approving takes one click or one keystroke. Rejecting returns the action to the agent with your reasoning, which becomes feedback for tuning the rules. Because evidence travels with every decision, an approval made in the Command Centre is still defensible months later — you can show exactly what was proposed, what it was based on, and who approved it.
Every agent action writes to an audit trail: who or what acted, what changed, which evidence supported it, and when. Operators filter that trail by tenant, workflow or timeframe, replay any action\u2019s context, and open a per-tenant drill-down showing lane activity, open approvals and recent corrections in one view.
The audit trail is the memory underneath the lanes. Every action — automated or human — records the actor, the change, the evidence and the timestamp, in order. Operators use it three ways: live monitoring while workflows run, per-tenant drill-down when a client asks what happened last week, and replay when something looks wrong. Because the trail is queryable rather than a log file you grep, governance reviews become filtering exercises instead of investigations.
Operators running one store use the Command Centre as their morning screen: open approvals first, workflow health second, audit exceptions third. If nothing needs a decision, the read takes minutes and the day belongs to the roadmap, not the queue. If you have been looking for a command centre ecommerce operators actually enjoy supervising, the design brief was simple: one screen, honest evidence, fast decisions.
Agencies running dozens of tenants get the more dramatic payoff. The console is tenant-safe by design: each client store is an isolated workspace with its own rules, approvals and audit trail, yet one operator supervises all of them from a single screen. Per-tenant drill-down shows lane activity, open approvals and recent corrections for that tenant only — no cross-tenant data bleed, no shared state, no risk that one client's pricing rule ever touches another client's catalogue.
The platform defines 17 agent roles. A public-safe subset shows the shape of the roster: a catalogue sync agent, a product content agent, a support drafting agent, a price-watch agent and a shelf audit agent. Each drafts or executes work under your rules, and each surfaces its output — and its evidence — in the relevant lane.
These are roles, not a shadow workforce. Each agent is scoped to a workflow, runs under your rules, and stops at its gate. The catalogue sync agent keeps channels consistent; the product content agent drafts and checks copy; the support drafting agent prepares replies with order context; the price-watch agent flags margin-risk changes; the shelf audit agent runs ShelfCheck's listing-rule checks. Whatever they do lands in a lane, with a trail behind it.
3 lanes
commerce_command, shelf_check and commerce_stack supervised in one console
17 roles
defined agent roles orchestrating under operator-set rules
100%
of high-risk actions gated by named-human approval
1
audit trail covering every lane, tenant and action
The Command Centre UI is deliberately read-heavy. Operators spend most of their time scanning state — what ran, what drifted, what needs a decision — and the interface is built for exactly that rhythm: dense information, quiet defaults, explicit approval moments. You do not configure workflows from the console; you supervise them.
Where input is needed, it is keyboard-first. Approve, reject, next approval, switch tenant, jump to lane — the common operator loop is reachable without touching the mouse. That matters at 8am with nine tenants and a coffee in one hand: the faster the loop, the more comfortably one person supervises governed automation across an entire portfolio.
The console opens read-heavy: lane status for the selected tenant, counts of open approvals, and anything the audit trail has flagged since your last visit. Keyboard navigation moves between lanes and tenants without a mouse.
The approval queue is ordered by risk and age. Each entry shows the proposed action, the evidence the AI used — the diff, the source data, the triggering rule — and exactly what happens if it ships.
Approve and the action executes and logs. Reject and it returns to the agent with your reasoning attached as feedback for tuning the rules. Either way, the decision is attributable to a named human from this moment on.
Because every action and decision lands in the audit trail, nothing needs remembering. Per-tenant drill-down reconstructs any day, any workflow, any change — for you, for a client, or for an auditor.
It is the operator console where a commerce team monitors automated work, approves high-risk actions and audits changes across every store and channel. BlueGreen’s Command Centre presents three lanes — Commerce Command, ShelfCheck and Commerce Stack — so one person supervises governed automation for one store or dozens of tenants from a single screen.
Four things, deliberately: monitor what agents did, approve or reject high-risk actions with the evidence attached, drill down per tenant, and read the audit trail for any action or timeframe. The console is read-heavy by design — it exists for decisions and supervision, not for data entry or workflow configuration.
You define what counts as high-risk: price changes above a threshold, bulk edits past a SKU limit, refunds, anything touching customer data. Those actions queue in the Command Centre with the diff, the underlying data and the triggering rule attached. Nothing ships until a named human approves it, and every decision is logged.
Yes — that is the design centre. The Command Centre is tenant-safe by design: every tenant store is an isolated workspace with its own rules, approvals and audit trail. One operator moves between tenants with a keystroke and supervises the whole portfolio from one screen, with per-tenant drill-down keeping each view cleanly scoped.
No. The console exists precisely because governed automation needs a human gate. Low-risk work proceeds automatically under your rules, but high-risk actions are always queued for human approval, every action is attributable, and the audit trail can be replayed. The promise is AI that acts with evidence — not AI that acts alone.
It is the operator surface of the BlueGreen governed commerce operating system rather than a standalone purchase. The lanes, the approval queue and the audit views come with running the underlying workflows. See the pricing page for how plans scale from a single store to an agency portfolio.
Book a 25-minute demo. We will connect a sample tenant, run governed workflows in all three lanes, and hand you the approval queue — evidence attached — so you can feel the operator loop yourself.