Conversation Analysis

Paste a chat transcript. We'll score it for buying intent, pull the topics, and tell you the next step you should take. This is exactly what Replyflow does on every conversation a visitor has on your site.

How buying-intent scoring works

Not every chat is worth a sales rep's time. Most aren't. The job of intent scoring is to surface the 5-10% that are, fast, so they get the attention they deserve before the visitor moves on.

The 0–100 scale

0–30 (Cold): information-seeking, FAQ-style, unrelated. No follow-up needed.
31–60 (Warm): real interest but no commitment signal yet. Newsletter, light nurture.
61–85 (Hot): explicit buying language, comparing options, asking pricing or security. Sales rep within 24h.
86–100 (Very hot): shared a corporate email, asked to talk to sales, named a deadline or seat count. Reach out within an hour.

What signals matter

Pricing questions are the strongest single signal. Mentions of team size, deadlines, or specific integration needs come next. Asking about security/compliance is a sleeper signal — it usually means the buyer is doing real evaluation, not browsing.

Where this falls apart

Short transcripts (1-2 messages) don't have enough signal — score them with caution. Visitors who already know what they want and just need a price don't tip their hand verbally; they're a 95 even if they only sent five words. The model accounts for that, but a rule-based scorer wouldn't.

Run this on every conversation, in real time

Replyflow scores every chat as it ends, tags topics, captures the email, and pushes the whole thing — transcript and all — into HubSpot. 500 free AI credits, no card.