💬 Governed, stateful semantic discussion
Your website knows more than it shows.
DISCOVER → DISCUSS → TRANSACT
DISCUSS is the conversation stage of the VISEON interaction journey: helping people and AI agents explore your offerings, ask follow-up questions and evaluate their options before taking action. It brings together Ask for people and MCP for AI agents, drawing on the same governed knowledge graph.
Ask provides a conversational question-and-answer interface that retains context, so users can ask follow-up questions naturally. Answers draw on your authoritative knowledge and supporting evidence. Clear to humans. Inspectable by machines.
Ask Premium adds a tailored personality and role behaviour, reflecting your organisation’s culture, training and working practices—for sales, product advice or support.
THE PROBLEM
What site search and generic chatbots cannot govern
Keyword search returns pages. Generic chatbots generate plausible prose. Neither inherently knows which source is authoritative, what context should persist, or when AI should not be called.
Links Without Answers
Traditional search matches terms and returns pages. The visitor still has to find, compare, and interpret the evidence.
AI Before Authority
Many conversational systems send every question directly to a language model, asking it to infer what should have been resolved from governed data.
Every Turn Starts Again
Stateless interfaces lose the subject, the chosen route, and the evidence already established. Follow-up questions become fresh guesses.
THE SOLUTION
How DISCUSS answers
Ask Naturally
The visitor asks in ordinary language. The current turn and committed conversation state enter the runtime together.
The Solution Routes the Turn
The Ask solution applies explicit rules, context, and source precedence to choose the correct branch: an exact entity answer, a relationship answer, evidence retrieval, clarification, or a bounded AI response.
Evidence Before AI
Known facts and relationships can be returned deterministically with no AI call. Where explanation is needed, DISCUSS retrieves governed evidence first and uses AI only to synthesise or narrate it.
Return the Answer and Its Provenance
The response can carry its evidence documents, graph nodes, and relationship identifiers, while the committed state remains available for the next question.
DETERMINISTIC → SEMANTIC → AI
The right authority answers each question
DISCUSS does not send every question straight to an LLM. It follows a governed order of authority, preserving state and exposing the evidence behind the answer.
Intelligent Semantic Routing
The solution governs the decision path. It decides what the question is asking, which source owns the answer, and whether AI is needed at all.
Stateful
Committed context, route selections, evidence, and interaction state are carried forward explicitly, so follow-up turns continue the same governed task.
Exact Graph Answers
The KG Skeleton and knowledge graph resolve known entities and relationships directly. Exact answers can be emitted without an LLM call.
GraphRAG Evidence Retrieval
When a question spans articles, service pages, FAQs, or case studies, DISCUSS retrieves semantically relevant evidence connected to the graph.
Bounded AI Synthesis
AI is used for explanation, composition, or voice only after the runtime has selected the branch and assembled the permitted evidence.
Provenance and Knowledge Gaps
Responses can expose evidence documents, graph node IDs, and relationship IDs. Unsupported questions are surfaced as gaps instead of being filled with invented certainty.
ONE GOVERNED INTERFACE. TWO AUDIENCES.
Stateful for people. Structured for machines.
FOR HUMAN VISITORS
Natural Conversation
Visitors ask questions in ordinary language rather than translating their needs into keywords or filters.
Context That Survives Follow-ups
DISCUSS retains the active subject, route, and established evidence as explicit state, rather than guessing from prose alone.
Answers With Evidence
Visitors receive an answer rather than a link list, with visible attribution and an honest boundary when the available evidence is insufficient.
FOR AI AGENTS & CRAWLERS
Queryable Semantic Contract
Agents can work with explicit entities, relationships, evidence, and response contracts rather than scraping meaning from rendered pages.
Machine-readable Provenance
Graph nodes, relationship identifiers, and source documents can travel with the answer, making its basis inspectable by downstream systems.
From Discussion to Action
The same governed context can support discovery, discussion, and later action. Authentication and RBAC remain authoritative in the underlying systems.
THE PREREQUISITE
DISCUSS performs in proportion to the governed context behind it
A sparse, contradictory, or unvalidated catalogue forces uncertainty. A governed knowledge graph, evidence index, and runtime contracts let DISCUSS know what it can answer exactly, what requires synthesis, and what it must not claim.
Digital Catalogue Semantic Entity Assessment
A structured evaluation of Schema coverage, entity relationships, source authority, content evidence, and the real questions DISCUSS must be able to answer.
Governed Context Build & Validation
VISEON builds and validates the Schema.org knowledge graph, KG Skeleton, GraphRAG evidence layer, and the machine-readable contracts that govern the runtime.
DISCUSS Runtime Deployment
A lightweight interface connects to the edge runtime, governed semantic sources, and evidence services. It does not require you to replace your CMS or the security model of connected source systems.
FAQ
Frequently asked questions
A standard chatbot usually sends each prompt directly to a language model. DISCUSS first routes the turn through an intelligent router, preserves committed state, prefers exact graph or KV answers, retrieves evidence when needed, and uses AI only for bounded synthesis or narration. The response can carry explicit provenance, and insufficient evidence remains visible.
No. GraphRAG is one evidence path inside DISCUSS; it is not the controller. The intelligent router owns routing and source precedence. Known facts and relationships can bypass GraphRAG and the LLM entirely, while broader questions can retrieve graph-connected content for bounded synthesis.
Yes, within explicit governed state. Conversation context carries the active subject and established evidence forward so follow-up questions can continue naturally.
DISCUSS needs a governed semantic foundation. A pilot can begin with a limited, validated domain, but production performance depends on reliable entities, relationships, evidence, and source authority. The Digital Catalogue Semantic Entity Assessment identifies what is ready and what must be built.
DISCUSS embeds through a lightweight web component connected to the VISEON edge runtime, knowledge graph, evidence layer, and state contracts. It is compatible with WordPress and other common CMS platforms. Where enterprise systems are connected, their existing authentication and RBAC controls remain authoritative.
Yes. DISCUSS can be scoped to one site or configured against a federated semantic catalogue covering a parent organisation, subsidiary brands, products, or services. Intelligent routing still applies explicit source and branch rules rather than treating every source as equally authoritative.
Using an MCP DISCUSS exposes governed entities, relationships, state, evidence, and provenance through machine-readable contracts. This gives agents a semantic interface they can query without treating rendered HTML or an LLM response as the source of truth.
Ready to let your website answer with authority?
DISCUSS makes governed context visible and usable: exact when the answer is known, evidence-led when synthesis is required, and explicit when the answer is not supported.