Order Agent

The Order Agent assembles a suggested order for every outlet, on every channel, from the same source of truth. It reads the outlet's reorder cadence, its recent sell-through, the last rep visit and the most recent shelf scan, then proposes a basket — line by line, with quantities — that reflects what the store actually moves rather than a generic template. When a field rep opens the mobile app, the order is already there to adjust. When the outlet logs into the B2B portal at midnight, the same basket is waiting. When a tele-sales agent picks up the call, they see the identical suggestion. One order, every channel, never out of sync and never duplicated.

Because the Order Agent runs on the ConnectX data layer — the single FMCG Cloud data model shared across the route to market — it builds orders that respect real-world constraints, not just past purchases. It can factor in active promotions, planogram gaps detected on the last shelf scan, void SKUs that should be reintroduced, and slow-movers that should be trimmed. A shelf finding becomes a reorder line; a promotion becomes a suggested add; a delisting becomes a removal. The agent does the merchandising thinking that a rep would do at their best, on every visit, for every outlet, whether or not a rep is in the room.

The Order Agent also closes its own loop. Every accepted, edited or rejected suggestion is written back to the same data model, so the next basket is sharper than the last — the agent learns which lines the buyer trusts, which it pads, and which it ignores. Those confirmed orders flow downstream into delivery routing and settlement on the same platform, so what gets suggested, what gets ordered and what gets delivered all reconcile against one record instead of three disconnected systems.