What you are about to watch
A two-minute freelancer chat. The assistant stays useful. Prism only speaks when the topic matches. Press Start demo. Input is off so every visitor sees the same four cards I do.
The scenario
Watch Prism insert partner tools inline. No banners, no popups.
Scripted playback. Direct input is off so the scenario stays consistent.
* Illustrative benchmark only.
AI Work Assistant
Scenario: freelancer planning the week
Live playback
Press Start demo to watch Jordan's conversation with inline partner ads.
Where the cards come from
Context in. Labelled offer or null out. No cookie. Matching waits until after the assistant replies, so the card follows the topic instead of interrupting it. Notes, drafting, shipping, a weekly plan. Sample partners: a runner, a restaurant, a florist, the runner again.
Related: the three lines on the homepage.
The same three lines
Pass the user message as context. Render a card only when getAd returns one. The API key stays on the server.
import { getAd } from "@prismpublication/sdk";
const offer = await getAd({
apiKey: process.env.PRISM_API_KEY,
format: "card",
context: { topic: userMessage },
});The script, in case you hate waiting
Playback types this. If you prefer reduced motion, read it here.
- User: buried in client notes across emails and docs. Bot: pick one tool and move everything there.
- Sponsored: Stride Classic Runner.
- User: project summary due this week. Bot: draft from the notes, then edit.
- Sponsored: Tokyo Sushi Restaurant.
- User: first-time sample shipping. Bot: prepaid labels, skip the monthly contract.
- Sponsored: Florista Flower Shop.
- User: put that into a weekly plan. Bot: notes today, draft tomorrow, ship before the samples leave.
- Sponsored: Stride Classic Runner (repeat sample).