Customer engagement · Live AI

Live personalization engine

A working look at the job an engagement platform does: a customer event fires, a profile is matched to a segment, and a personalized message is written for each channel. The copy below is generated live by an AI model running on the edge, not pulled from a script.

1. Pick a customer

2. Fire a trigger event

How it works

Every Generate runs a real request path, the same shape a production engagement platform uses.

  1. Step 1

    Pick a profile and an event. The customer is held as JSON, the kind of record a CDP would store. The event is the trigger that starts a journey.

  2. Step 2

    The browser posts it to the edge. A POST /api/personalize hits a Cloudflare Pages Function. No API key ever ships to the browser.

  3. Step 3

    Workers AI writes the message. The function prompts a Llama 3.3 70B model through the edge AI binding for strict JSON: a segment, a rationale, and one message per channel.

  4. Step 4

    The page renders the result. The three channel cards fill in. If the model is ever cold or rate-limited, a templated fallback keeps the demo working.

Architecture: a static Astro site, a serverless Pages Function, and edge AI, all on the free tier. There is no server to run.

Mapped to the marketing concepts

A concept demo of the engagement layer, not a third-party platform integration. The copy is generated live by a model running on Cloudflare Workers AI at the edge. Built while prepping for a Solutions Consulting conversation in the customer-engagement space.

More MarTech demos