Twelve months on the desk: 35% to 73%.

The desk before

The picture will be familiar: RFQs arriving in five formats across email and WhatsApp, a pricing team assembling quotes from rate cards and portal logins, sixty-plus minutes per response on a good day — and volume climbing past what the team could answer. Quotes that went out were solid; the problem was how many went out late, and how many never went out at all.

What changed

The quotation agent went live under supervised approval: it reads each inbound RFQ, prices it across Infinity’s rate sources, drafts the reply with margin checks, and holds at the gate for a coordinator’s tap. Acceptance and sending never left human control — that gate is why the team trusted it enough to keep dialing drafting toward auto.

In their words

The phrase worth noticing is perfect memory. The desk’s speed shows up in the metrics; its memory shows up in the quiet things — the surcharge line that never goes missing again, the customer pattern surfaced on every draft, the 2 a.m. RFQ answered before the competition’s office opened.

Where the story continues

Quoting was Infinity’s wedge, not the whole deployment: the same platform runs booking, tracking, reconciliation and collections, and payment history now feeds credit-aware quoting on the next RFQ. The full multi-agent story lives on the FreighAI case study; the collections chapter is told at Receivables AI.