1.65M emails turned into 1.65M conversations with context

UTN BA had a chatbot in development capturing lead conversations, but that signal was lost at the email layer: everyone got the same template — no nurturing, no funnel-stage segmentation.

The challenge wasn't volume — it was disconnection.

• The chatbot captured each lead's specific intent (which course, which doubt, which level).
• But email consumed that signal as a static template: everyone got the same message.
• No nurturing strategy, no funnel-stage segmentation.
• Result: flat engagement and rising CAC, with the most valuable signal — the conversation itself — sleeping in the system.

Previous attempts to process thousands of chat histories in real time collapsed the infrastructure: timeouts, latency, intermittent errors. This wasn't a data-availability problem — the chatbot, the CRM, and the data all existed. It was a scale problem: 351,000+ students, 650 courses, and roughly 1.65M sends per two-month cycle, run through a naive real-time lookup that couldn't hold under that load.

UTN BA didn't need another tool — they needed a technical partner to articulate the pieces they already had. From Cleon we designed four moves:

• An evolved conversational layer, built alongside UTN BA's own technical team.
• Automated, generative email anchored to the conversation ID — the central innovation.
• The chatbot repositioned as orchestrator across CRM, LLM, and Marketing Cloud.
• An always-on journey, replacing the concept of a "mass send" entirely.

The chatbot acts as the orchestrator — talking to the LLM through API + MCP. Marketing Cloud pulls leads directly from Salesforce and queries the orchestrator's API at send-time for each email's generative body.

Cleon worked with UTN BA's technical team to evolve the chatbot and introduced a new architecture: nurturing leads at different funnel stages and generating hyper-personalized emails dynamically through a language model, using the lead's conversation ID as the key to retrieve the full history.

Cleon's role: architecture, implementation, and operation.

Every chatbot conversation triggers an individualized flow — we removed the concept of "mass send" from the stack entirely.

Zero system downtime over 1.65M sends in its first two months live.

The stack: SFMC Journey Builder for always-on, per-lead journeys; Salesforce CRM holding Lead/Opportunity state; a language model reached via API + MCP (Model Context Protocol) generating each email's body; and the chatbot itself repositioned as the orchestrator tying the three together.

Email went from a declining channel to the highest-relative-engagement channel at UTN BA. A ~20 percentage-point lift in open rate on 1.65M sends translates into incremental qualified-lead pipeline that the institution absorbs without adding marketing headcount — the marginal cost of the next personalized email trends to zero.

The conversation ID → CRM → language model → Marketing Cloud architecture solved a problem most orgs with chatbot + CRM + email have today: the pieces exist, the layer that articulates them does not. From this pattern, Cleon developed an in-house leadbot + generative orchestration product that extends the same architecture to clients without an existing chatbot.

First two months post go-live (Jan–Feb 2026).