By the Onsites AI team · Last updated · 4-minute read
Buying intent arrives most often as a support question — "how much for 60 units to Rotterdam," asked in a chat, at odd hours, by someone who may never email at all. Most teams lose the deal in the seam: the agent writes down the request, promises a quote, and the handoff to another tool, another inbox and another day lets enthusiasm cool into silence. The chat-to-quote flow closes the seam: from inside the conversation, the agent drafts a real quote — Onsites Mail's editor, the shared product catalog, the money sentence — and sends it into the same thread the buyer is already watching, where it inherits the statuses and the account linkage automatically. This guide covers the spotting, the drafting, and the cross-border variant.
Chat buys when four signals stack, and they are teachable in an afternoon. Specificity: "40ft for 60 units, delivered to Rotterdam" is a buyer; "how much is shipping" is a browser — the quote's unit economics only exist for the specific order. Relationship context: the account panel shows prior orders — a third order asking price is a re-quote (different conversation, faster lane), a first-timer needs the trust-building terms the freight playbook handles. Urgency markers: "we need it by the 15th" is a production slot, not a curiosity — quotes with deadlines go out same-day or the slot goes to someone else. The question behind the question: "does it work for 60 units?" usually means "is the unit price right at 60?" — and the tier table is the actual answer. Agents trained on the four signals quote in the first session; teams without them "send it later" and lose the buyer to the seller who answered in-thread.
A declined quote is worth more than a silent one, and the status trail is how a small team learns pricing from its own history. Declines get a reason — logged once on the account, visible at the next quote: "price" teaches a tier gap ("at 100 units we drop 8% — worth a re-quote?"), "timing" teaches a calendar (this buyer buys in October), "went elsewhere" teaches the market's honest margin. The pattern compounds at the account level: the buyer who declined twice at one price and accepted once at another has taught you their real number; the one who declined a first order and never returned taught you where the volume ceiling of one product tier sits. This is also where AI helps most cheaply: the negotiation summary across versions, the decline reasons mined at the quarterly KPI review — while the human decides what the market just said. A desk that reads its declines quotes better within a season; one that treats them as statuses to file repeats the same negotiation, with the same buyers, at the same surprise, forever.
Two habits turn quotes from sent-and-hoped into a managed lane. Deliberate validity: price validity is a deal design, not boilerplate — long enough for the buyer's approval cycle (two weeks is a sane default for B2B, shorter for spot deals), explicit in the money sentence, and carrying the honest "prices may change after date X" if your inputs do. Quotes that expire silently teach buyers to shop elsewhere; quotes that announce their expiry re-open the conversation exactly once, on purpose. The expiry follow-up: when a quote lapses without acceptance, the desk's statuses fire the one-line check — "the Rotterdam quote expired Friday; is this still live on your side?" — which converts a large share of lapsed quotes into re-quotes and a few into declines that are at least known declines instead of mysteries. Both habits live on the same status trail: draft, sent, accepted, declined, expired, followed up — the account view reads the whole lane, and the "did we win this?" question, which eats more founder memory than any other, becomes a lookup.
The flow's mechanics matter less than its discipline: the quote must be drafted from the shared catalog — the same items, the same names, the same prices every other document quotes — so there is nothing to re-type and nothing to drift. The quote's money sentence carries the deal exactly once (amount, currency, terms, validity window); the notes field records what the buyer actually asked, in their words, so the acceptance conversation never negotiates a paraphrase. Sent quotes land in the thread — the buyer sees the price where they asked the question, and any follow-up ("can we do 70 units?") continues the same lineage rather than starting a new thread from memory. And the desk keeps the whole thing honest afterward: the status trail (draft, sent, accepted, declined, expired), the opportunity record that was quietly created when the intent appeared, and the account view showing the full relationship — the quote, the order it became, the invoice trail after.
Trade conversations add two layers, and the draft flow survives both. The proforma invoice: international buyers pay against a proforma — effectively the quote in invoice form (the full anatomy in the RFQ-to-proforma guide) — and it flows from the same chat thread through the same catalog, so the buyer's bank gets the document the buyer's chat already knows. The requote chain: foreign-trade buyers negotiate in spirals ("and for 500 more units? 40ft instead? payment 60 days?") — each turn is a new quote version on the same account lineage, so the thread shows the whole negotiation honestly, and the finally-accepted version is the one the order inherits. The copilot helps most in the chain: translating the buyer's requote (3 credits), summarizing what a five-turn negotiation agreed (6 credits), while the human edits and signs each version — because the money sentence still has a name on it at every step.
What is the chat-to-quote workflow?
Support chat detects buying intent — the buyer is recognized, the account linked, the request specific — and the quote is drafted from the shared product catalog without leaving the conversation window. Sent, it lives in the same thread as the chat that created it, inheriting the account's statuses and history.
How do you spot buying intent in support chat?
Four teachable signals: specificity (units, container type, delivery city — quotes need real order shapes), relationship context (a repeat buyer pricing a re-order vs a first-timer needing trust-building), urgency markers (deadlines are production slots), and the question behind the question (unit price at volume). Teach the list for an afternoon; quote in-session.
Why draft the quote from a shared catalog?
Because re-typed prices drift: the quote, the order and the invoice must carry the same numbers, from one catalog updated in one place. The chat thread keeps the buyer's words; the account lineage keeps the negotiation's history honest.
What changes for cross-border sellers?
Two additions on the same flow: the proforma invoice (the quote international buyers pay against) and the requote chain — every "and for 500 more?" is a new version on the same account lineage, so the negotiation stays one auditable story and the accepted version is exactly what fulfillment works from.
Where does the CRM fit in?
The buyer was already linked the moment they wrote in with intent; the opportunity record tracks the negotiation; the quote's statuses (draft → sent → accepted) live on the account's story — so "did we win this?" is answered by the record, not the memory.