Answering half the support queue without hiring

BigBuy's B2B customers are dropshippers and resellers running businesses on top of its catalogue. They generated up to 200 new support tickets a day, in whatever language they spoke, and the only automation in place was a set of templates that, in the team's own words, did not always fit.

BigBot is an orchestrator routing to five specialist sub-agents: Commercial, Support, Products, Business Agreement and Orders. The topology copies the human team, because those divisions already existed inside support and already worked. Each sub-agent's description defines its negative boundaries too, so it knows what is not its job and who to hand it to. That is the difference between five agents and a routed system.

The orchestrator does more than route. Sub-agents return raw technical material with no formatting, and the orchestrator is the only thing that ever speaks to the customer: it owns tone, language and presentation. Separating who solves the problem from who says it is what stops five specialists from sounding like five different companies.

It does not only answer. It generates supply contracts through the API and attaches them, pulls proof of delivery, and moves each ticket to the right pipeline in HubSpot. The whole thing lives inside the CRM support reps already used, so nobody changed tool and results were visible from day one.

Two people built it. One engineer, and me. The cost argument that mattered was not headcount. Support was an entry-level role with heavy turnover, so knowledge kept walking out of the door. A system with a knowledge base and versioned prompts holds on to what turnover keeps taking away.