Guided Experiences
A bot that only answers when asked a question is a search box with extra steps. A guided experience shapes what a visitor sees before, during, and after they ask anything at all.
Welcome Screens and Multi-Panel Layouts
Rather than dropping a visitor into a blank chat box, a well-designed welcome screen offers quick actions, featured topics, forms, and guided journeys up front: multi-modal (text, voice, image upload), with automatic language detection and a mobile-friendly layout that can expand to full screen. Configure this by brand, audience, page, or campaign rather than shipping one generic welcome screen across an entire site.
Directives: AI-Controlled Page Experiences
Directives let the AI do more than answer in text: it can embed an invisible tag in its response that triggers a real-time action on the page: switch a visible panel, navigate to the most relevant URL, show a form, launch a video, load a calculator, or trigger a custom workflow. The directive tag itself is stripped from the visible answer, so the visitor just experiences a natural conversation and a page that adapts around them.
The important distinction to make with a client: this isn't hardcoded navigation rules. The AI reasons about the visitor's intent, then triggers the matching page behavior: ask about retirement planning on a wealth-management site and it can switch to the retirement panel without the visitor clicking anything.
Attributes: Feeding Context to the AI
Directives push actions out to the page. Attributes are the reverse: a dictionary of runtime data the client's website passes in to the AI as the conversation starts — page type, product category, logged-in state, language, CDP data, geolocation, whatever the client's stack already knows about the visitor. The developer sets it once (via a dataAttributes/{attributes} token in the system prompt); after that, the AI has that context on every message without the visitor having to state it.
The distinction to make with a client: Attributes and Directives are the two halves of a context-aware experience. Attributes tell the AI where the visitor is starting from; Directives are how the AI acts on what it learns from there. Attributes are what make Landing Page Experience Controls, below, possible.
Landing Page Experience Controls
Personalization doesn't have to wait for the visitor to type anything. Using Attributes — signals like CDP data, geolocation, or cookies passed in as the page loads — ai12z can personalize a landing page's content, interactive controls, recommendations, and even the bot's own welcome panels before the first message is sent, see Connecting CDP to ai12z and Intent + Persona Personalizer for how the underlying signals are configured. Once the visitor does start typing and reveals a different intent, Directives can update the page live: a landing page that opened personalized around one product line or region can shift to another in real time, without the visitor navigating away or starting over.
Vibe Coding
Describe the experience you want in plain language, and ai12z generates the layout, content, and interactive components (welcome screens, HTML widgets, forms, prompt improvements) without a full development cycle. Every change is saved in history, so a team can review and roll back at any time. See Vibe Coding for the full walkthrough. As covered in The ai12z Advantage, this is available at every engagement tier: it's how you iterate quickly regardless of how advanced the underlying deployment is.
Related Documentation
- Directives: Full technical reference
- Passing Attributes and Tags: Full technical reference
- Vibe Coding: Full technical reference
- Intent + Persona Personalizer: The persona-detection layer behind Landing Page Experience Controls
- Engagement Tiers: Where guided experiences fit in the capability matrix