From Ad Click to Enrolment in One Conversation: Designing an Agentic AI Journey for Cruz Roja
Agentic AI · Conversational AI · RAG · Nonprofit · Digital Ecosystems
Most organisations do not have a lead problem. They have a lead-waiting problem.
A person sees an ad for a first-aid course, taps it, fills in a form, and then waits. Someone calls back the next morning or the next week. By then the moment of intent has passed.
The research on this is old and still uncomfortable. The MIT and Inside-Sales lead response study found that reaching a lead within five minutes made contact about 100 times more likely, and qualification about 21 times more likely, than reaching them at 30 minutes (summary). Few human teams can respond in five minutes at 11 pm on a Sunday. An agent can.
The brief: many offerings, many countries, one front door
Innovun Global designed an agentic AI solution for Red Cross across the USA and Latin America. The challenge was not one product but three very different ones:
• Short-term courses (one-day programmes) sold through a Shopify storefront.
• Long-term courses with payments every six months, run on a subscription model.
• Ambulance booking, where speed and accuracy matter more than persuasion.
Each had its own audience, price logic, and urgency. A single contact form could not serve all three well.
The architecture: WhatsApp first, website second, one brain behind both
We designed the journey in two phases.
Phase 1—a centralized WhatsApp agent. Google, Facebook, and Instagram campaigns all route into one WhatsApp agent. In much of Latin America, WhatsApp is where people already talk to businesses, so the agent meets them there instead of sending them to a form.
Phase 2 — the website and a RAG agent. The Cruz Roja website gets its own agent, drawing on the same knowledge as WhatsApp. Behind both sits a vector database and a large language model. Retrieval-augmented generation (RAG) means the agent answers from Cruz Roja's own approved course and service content, not from the open internet.
Every conversation follows the same five-step journey:
Greet — recognise the channel, the campaign and the language.
Qualify — understand which offering fits: a one-day course, a long-term programme or an urgent service.
Educate — answer questions on content, schedule, certification and price, grounded in approved material.
Propose — recommend the specific course or service and the right payment plan.
Confirm — take payment inside the flow by card, UPI or Spin, and confirm the booking.
Every lead lands in a lead database and a Smart Lead Management Dashboard. Where a human is needed, the agent hands off to regional SPOCs or the academic team with the full conversation attached, so nobody asks the same question twice.
Four design principles we would apply anywhere
• Ground the agent, do not unleash it. For a humanitarian brand, one wrong answer about a certification or an ambulance is a trust problem. RAG keeps the agent inside approved content.
• Design the handoff before the automation. The most important rule an agent follows is when to stop and bring in a person. Urgent services skip persuasion and go straight to action.
• One journey, many offerings. Greet, Qualify, Educate, Propose, Confirm works for a one-day course and a six-month programme alike. Only the content changes.
• Close the loop in the dashboard. Leadership sees campaign to conversation to payment in one view, which is what turns ad spend into an accountable number.
Why this matters beyond nonprofits
Schools, hospitals, training institutes and service brands share the same pattern: high-intent leads, many offerings and small teams. An agent that answers in seconds, in the customer's language and on the customer's app, does not replace those teams. It gives them back the leads they were losing while they slept.
Want to see the Greet-to-Confirm journey mapped onto your own funnel? Book a session with Innovun Global.
Comments