Knowledge Assistant — Product Walkthrough (Evaluation Release)

Overview

This walkthrough shows Knowledge Assistant in the Schema App evaluation release (~7 minutes). You will see three patterns that matter for enterprise evals:

1. Cross-site synthesis — one answer from marketing and support, with citations

2. FAQ rich results — FAQPage markup rendered as an accordion

3. Contact routing — ContactPoint entities as actionable contact cards

The recording also briefly covers conversational memory, vector-based semantic search, and a preview of the agentic web / WebMCP direction.

Recording quality: This is an evaluation-quality screen capture intended for support and customer onboarding — not a marketing-polished production video.

Recorded: June 17, 2026 · Presenter: Mark van Berkel, CTO

Related articles:

• Knowledge Assistant — Evaluation Release Guide — written evaluation steps, six test scenarios, troubleshooting, and scope details

• Knowledge Assistant — Product Walkthrough (Evaluation Release) — this article (~7 min video)

Watch the walkthrough

Before you watch (or try it yourself)

The demo uses Schema App’s internal Knowledge Assistant at https://app.schemaapp.com/knowledge-assistant. To reproduce the same patterns on your tenant:

• Open Knowledge Assistant and click Clear History

• Set the site filter to All sites (required for cross-site demo)

• Set the relevance threshold slider toward the lower end if contact cards do not appear on contact-intent questions

You need evaluation access enabled for your account. If Knowledge Assistant is not visible, contact your CSM.

Chapter guide

Try these prompts in your evaluation

After watching, run the same question types against your markup. Substitute Schema App examples with your products, FAQs, and contact pages.

1. Cross-site synthesis

What is Entity Hub and how do I perform an Entity Audit?

Look for: Overview and Key Facts; citations from more than one site or subdomain when All sites is selected.

2. FAQ rich result

How does the Microdata Filter work?

Look for: An FAQ accordion sourced from FAQPage markup. If you only see a short paragraph, confirm FAQPage coverage and try lowering the relevance slider.

3. Contact routing

What is the contact for sales inquiries?

Look for: ContactPoint routing cards with email, phone, or URL — each tied to a cited source page.

What this walkthrough does not cover

The video focuses on conversational UI patterns in the evaluation release. It does not walk through:

• Hosted deployment on your public website

• Knowledge Assistant Settings and branding configuration (see the Evaluation Release Guide)

• Tenant MCP endpoint setup (contact your CSM)

• Booking, registration, or other action workflows (roadmap)

Summary transcript (edited)

Below is an edited summary of the recording for search and accessibility. Timestamps are approximate.

0:00 — Knowledge Assistant and citations

Knowledge Assistant is Schema App’s NLWeb-based conversational interface for your Content Knowledge Graph. It uses structured schema.org data you already curate. Every answer includes citations — the entity and the source page — so responses stay grounded in your content, not the open web.

0:36 — Cross-site synthesis

The demo queries across multiple Schema App properties (marketing, support, and related sites) with the site filter set to include all enabled sites. A question about Entity Hub and entity audits returns one synthesized answer with overview, key facts, and clickable sources spanning more than one subdomain — one governed graph, multiple surfaces. Administrators can control which sites are in scope when deploying to a website.

2:05 — FAQ rich results

When the best source is an FAQPage, Knowledge Assistant reuses your existing FAQ question-and-answer markup and renders an accordion — surfacing related questions people might ask, not a single invented paragraph.

3:00 — Contact point routing

ContactPoint markup surfaces as routing cards broken down by channel or department (for example sales vs customer support vs general inquiries), similar in spirit to contact rich results in search. This pattern is especially useful for large organizations with many divisions.

3:55 — Preview and semantic search

The in-app experience is a preview of how conversational search can work on your site — moving from lexical keyword matching to vector-based semantic search over schema data. Branding and administrative controls apply when you go to a hosted experience.

4:51 — Conversation memory

Users can refine questions in a thread; conversation history persists in the browser during evaluation so you can return and continue (client-side memory in the evaluation release).

5:10 — Agentic web and WebMCP

The same governed graph can serve AI agents that discover and invoke tools on your site — not only human chat. The recording includes a brief WebMCP preview (early Chrome preview; broader industry support in progress) showing how browser agents may query structured site content. Action workflows such as bookings or resource requests are part of the longer-term agentic web roadmap, not the current evaluation release.

Related articles

• Knowledge Assistant — Evaluation Release Guide — work through all six recommended test scenarios after watching

Next steps

1. Read the Evaluation Release Guide and work through all six recommended test scenarios

2. Note coverage gaps — thin answers are a markup backlog, not a failure

3. Contact your CSM to discuss hosted deployment, MCP endpoints, or public rollout when you are ready

Questions? Contact your Customer Success Manager or submit a support request through Schema App.