korab eland
← all case studies
case study in flight

Chat was never built to capture leads. It was capturing them anyway.

Chat was a support service, not a lead channel. But a third of visitors volunteered their email, and almost none reached the CRM. Transcript analysis at scale sized the genuinely incremental opportunity honestly, and an automation bridge now captures it.

role
Led analysis + integration design
team
Growth, IT, marketing automation
stack
Python · AI transcript analysis · Dynamics 365
span
2026
status
in flight

Our lead capture ran through the form. Chat was a support service, there to answer questions, never wired to create leads. Yet over 13 months around 64,000 conversations passed through it, and about a third of visitors typed in their email. About 25 a month became CRM leads, logged by hand when an agent judged a conversation worth it. That low number was by design, not neglect. What nobody had done was analyse the transcripts at scale to see what the channel was quietly holding: a stream of high-intent contacts the business had never counted.

The obvious headline was also misleading, and saying so mattered. Manually captured chat leads converted exceptionally well, but they were high-intent by construction, so extrapolating their rates would have overstated the prize. I cross-referenced chat emails against existing CRM records, found around 40% were already known via brochure downloads, and modelled only the genuinely incremental remainder at conservative, mid and optimistic conversion rates: roughly 200 to 1,000 incremental registrations a year. The business case led with the honest number, and it still cleared easily. The integration pays for itself with fewer than two additional registrations a year.

The fix was deliberately boring: a lightweight automation bridge from the chat platform into Dynamics 365, plus an engagement strategy for email-only leads the outbound call model could not reach. The first automated journey now performs at parity with the manual process it replaced, and a channel nobody used to count is finally feeding the pipeline.

outcome

Chat-provided emails that never became CRM leads ~98% — how this was measured

Thirteen months of chat transcripts, around 64,000 conversations, cross-referenced against CRM lead records. Roughly 1,600 email-providing chats a month against about 25 manually captured leads.

Modelled incremental registrations a year ~200–1,000 — how this was measured

Scenario model at conservative, mid and optimistic conversion rates for the genuinely incremental share of chat leads, after removing the ~40% already in the CRM via brochure downloads.