01Generic widget
Duplicate knowledge
A separate widget answers from copied snippets while the real website and CMS keep changing underneath it.
A chatbot that answers with your business context.
Why generic chatbots fall short
Generic chatbots fail when they sit outside the business.
A separate FAQ tool means duplicate knowledge, disconnected transcripts, weak Relationship Intelligence context, and no real loop for fixing what it gets wrong.
02Generic widget
Lost context
A useful sales question becomes a transcript in another tool instead of Relationship Intelligence context the team can act on.
03Generic widget
Weak review loop
Bad answers disappear into logs instead of becoming knowledge gaps the team can actually fix.
04Generic widget
No real boundary
The widget either does too little, or starts implying actions the business never agreed to review.
Inside Built-in Chatbot
Chat shares the same context as your site, Relationship Intelligence, and content.
Tej deploys the chat foundation with every project. Its knowledge, Relationship Intelligence access, engagement rules, handoff, and approved actions are configured around the platform.
Approved knowledge
Pages, posts, static content, Q&A, documents, and AI Engine sources define what the chatbot is allowed to answer from.
knowledge sources
Relationship Intelligence context
When identity and permissions allow, known visitors can carry contact history, lead status, activity, and previous conversations into chat.
identity + history
Controlled engagement
Prompts can respond to page, time, scroll, return visit, intent, and lead score, with caps and cooldowns.
rules + caps
Human handoff
Escalation, admin replies, AI pause states, and notifications are part of the workflow, not an afterthought.
escalate + pause
Confirmed actions
Booking, contact capture, routing, and qualification can be configured as approved workflows with confirmation, logs, and guardrails.
approve + log
Knowledge gaps
Weak answers and unhelpful feedback can become reviewable improvements for content and source coverage.
feedback loop
Knowledge grounding
The chat layer should know where its answer came from.
The strongest chat systems are not more autonomous. They are more grounded, more observable, and easier to correct when the content is missing.
01Step
Approved sources
Editors choose the pages, posts, Q&A, static content, and uploaded documents the chatbot can use.
02Step
Retrieval through AI Engine
The chat layer retrieves relevant context from the project knowledge base instead of relying on a blank prompt.
03Step
Grounded response
The response uses retrieved context, brand guidance, permitted Relationship Intelligence state, and the project’s configured action rules.
04Step
Reviewable gap
Low confidence, negative feedback, and missing coverage can become knowledge gaps for the team to inspect.
retrieval sample
Visitor asks about implementation. Context retrieved from the services page, the Relationship Intelligence build brief, and the approved project-call workflow.
Freshness
Source updates can flow from the CMS and knowledge base.
Gap capture
Weak answers can become review tasks instead of hidden failures.
Handoff and review
A useful chatbot knows when to stop.
When actions are configured for the project, the chat layer can help qualify, route, book, or draft. Each action keeps a clear boundary and a human-visible record.
operator consolereview required
Escalate
Offer handoff when confidence is low or the visitor asks for a person.
Pause AI
Let an admin take over without the chatbot continuing in parallel.
Confirm action
Booking, contact creation, and routing stay explicit and logged.
Export
Give the team transcripts and summaries for review or handoff.
Approved actions
Booking flowFollow-up draftGuardrail check
Actions are scoped to the workflows configured for this project.