How Live Web Agent Works

Live Web Agent uses advanced AI to read your website, understand what visitors are asking, and deliver accurate answers in real time. When it can’t help, it hands off to your human team. Here’s how that works.

STEP 1

Website Content Ingestion

When you connect your WordPress site, Live Web Agent automatically reads and indexes your content:

• Pages and posts: Every published page and blog post gets fetched and processed.
• Documents: Upload PDFs, spec sheets, and manuals directly.
• Automatic sync: When you publish or update content in WordPress, it re-indexes instantly. No manual triggers needed.
• Smart chunking: Content is broken into semantic chunks that preserve meaning, so retrieval stays accurate even on long pages.

No manual tagging. No keyword lists to maintain. Connect it and you’re done.

STEP 2

A Visitor Asks a Question

When a visitor opens the chat widget and types a question:

1. Understanding intent: The system understands the question by meaning, not just keywords.
2. Semantic retrieval: Finds the most relevant content chunks by actual meaning, not keyword matching.
3. Keyword boosting: For queries with specific product names or model numbers, a keyword search layer runs in parallel and injects exact matches the semantic search might miss.
4. Reranking: Results are reranked for relevance, so the best information gets prioritized.

The dual retrieval approach (semantic + keyword) lets the agent handle fuzzy questions like “What’s your most affordable option for a small team?” and precise ones like “Does the NR2110-O come in stainless steel?” with the same accuracy.

STEP 3

Answers are Generated

The retrieved content is passed to the AI model along with your custom system prompt:

• Strict mode: The agent only answers based on retrieved content. No outside knowledge, no fabrication.
• Anti-hallucination guardrails: Hard rules prevent inventing product numbers, specs, pricing, or URLs that don’t exist in your content.
• Brand voice: Your custom system prompt controls tone, style, and behavior.
• Natural responses: No “based on the provided context” hedging. The agent speaks confidently and naturally.

STEP 4

How it Escalates to Humans (Live Agents)

The agent knows when to hand off:

• Explicit requests: If a visitor asks to speak with a human, the agent immediately escalates.
• Confidence gaps: When the agent has no relevant information to answer, it escalates instead of guessing.
• Seamless handoff: The human agent receives the full conversation history, so the visitor never has to repeat themselves.
• Configurable timeout: Set how long to wait for a human agent before showing a contact form.
• Real-time alerts: On-site agents receive instant notifications when a visitor needs help.

STEP 5: WHAT SETS US APART

The Gap Agent: 24/7 Continuous Improvement

Here’s what the Gap Agent does:

1. Gap detection: The Gap Agent analyzes your content and visitor conversations to find topics people ask about that your site doesn’t cover.
2. Prioritization: Gaps are ranked by frequency and severity. The most-asked unanswered questions surface first.
3. Content drafting: It can generate draft content to fill gaps, using your existing content as context.
4. Clarification workflow: When it identifies a gap, the agent can ask the visitor clarifying questions to better understand what they need, capturing valuable intent data.
5. Publishing pipeline: Draft content goes through your review before publishing.

The result: fewer unanswered questions over time, better answer quality, and a content strategy driven by what your visitors actually want to know.

Google Analytics Integration

Live Web Agent pushes chat events directly into your GA4 Analytics through the Google Tag Manager:

• Every time the Chat widget opened
• When Messages are sent
• Escalations to human agents
• Lead capture even when a form was submitted

Everything shows up alongside your existing website analytics, no separate dashboards required.