Case studies / B2B agricultural export — sales operations
1,824 buyer enquiries from email and WhatsApp, deduplicated, scored and moved through one pipeline
Since April 2026 an exporter's enquiries from email and WhatsApp have landed in one pipeline instead of two inboxes and a phone. 1,824 deduplicated enquiries have been captured; each is scored for genuineness, enriched with what is known about the buyer, and moved through stages — 482 contacted, 67 negotiating, 9 quoted.
Radiance Overseas · India, selling internationally
Built and run in-house — this is our own property, not a client engagement, which is why every number can be produced on request.
- 1,824
- enquiries captured, deduplicated
- 385
- arrived on WhatsApp, not email
- 482
- contacted
- 67
- in negotiation
Source: Production CRM, deduplicated rows, 16 Apr – 12 Oct 2026.
What was the problem?
Export enquiries arrive everywhere: contact forms, marketplace emails, forwarded threads, WhatsApp messages to a personal number. The same buyer writes twice. Spam and fishing enquiries look exactly like real ones. Nobody can say how many enquiries came in last month, let alone which were answered.
Every enquiry that goes unanswered for a day is a buyer who has already asked three other suppliers.
What did we build?
One intake for every channel
Email and WhatsApp feed the same pipeline. Duplicates are detected and merged, so one buyer is one record however many times they write.
AI where judgement helps — and nowhere near the numbers
A model reads each enquiry and scores how genuine it looks, and what is known about the buyer is gathered alongside it. A reply draft is one button away, but a person reads it and sends it — the model never sets a price or sends anything on its own.
A reply queue that cannot quietly forget
Every enquiry has a stage — new, contacted, negotiating, quoted — and a revised bid goes back into the 24-hour reply queue rather than sitting in a thread.
What changed?
- 1,824 deduplicated enquiries captured since 16 April 2026: 1,439 by email and 385 by WhatsApp — a fifth of all demand that previously lived on one phone.
- 482 contacted, 67 in negotiation and 9 quoted, every one traceable to the message that started it.
- It runs on the business's own infrastructure; there is no per-seat CRM subscription.
What is not working yet?
1,264 enquiries are still at “new”. Capture and triage were the first problem and they are solved; working the backlog is a staffing question that software can make faster but cannot answer. We show the stage counts as they are rather than only the flattering end.
Common questions
Can an AI agent answer customer enquiries automatically?
It can draft them. We do not let it send on its own where quotes and prices are involved — a wrong number in an export quote is expensive. Here the model scores and drafts on request, and a person sends.
Do I need a CRM like HubSpot for this?
Not for capture, deduplication and a reply pipeline. This one runs on the business's own server with no per-seat fee. What cannot be self-hosted is rented data — contact databases and market feeds are still bought.
Could this work for your business?
It probably fits if:
- Enquiries reach you by email and WhatsApp and nobody sees the total
- You suspect leads are getting lost but cannot prove it
- You pay per seat for a CRM your team does not keep up to date
What this used
More case studies
Student accommodation
From invisible to 577 search clicks a quarter — and recommended by ChatGPT in 73% of student questions
38.5× search clicks, 15 → 577 per 90 days
Student accommodation — paid search
The Google Ads dashboard said ₹243 a lead. The real figure was ₹1,100. Here is what we found and fixed.
~40% of recorded conversions were fraud
B2B agricultural export
856 indexed product pages for a B2B exporter — and named by ChatGPT and Gemini to international buyers
6.6× search clicks, 58 → 385 per 90 days