VISEON for Qlik · vCat
Your AI should understand what you mean before it computes the answer.
vCat is the top-level navigation system for enterprise AI in your Qlik Cloud Analytics tenant. It keeps the canonical master map of your tenant (your live reporting, master items and operating rules) so AI agents know and understand your estate before they act, and every answer carries its governed definition and trust level.
The problem
The gap data access doesn’t close
Data access tells an AI agent what exists. It doesn’t tell the agent what things mean. The risk is AI running your reporting and operations without your business context.
The semantic gap
Where meaning must be established: largely unserved today- 1Interpret intentRead the question and find the business goal behind it.
- 2Resolve business meaningSettle what the business terms actually mean.
- 3Identify the metricPick the metric that answers the question.
Data access
What existing tools already do well- 4Select the implementationChoose the logic that computes it.
- 5Execute the analysisRun the calculation and return the output.
The challenge is not retrieving data. It is establishing the right semantic path before reasoning over it.
Ask an AI agent a business question and it travels this path. Tools that give agents data access, such as a Qlik MCP server or a SQL agent, are good at the last two steps. They are not built for the first three: understanding intent, resolving meaning and picking the metric that actually answers the question. That’s the semantic gap, and it’s where most AI-on-analytics projects go wrong: not at the compute, but before it.
How it works
From your analytics estate to governed concepts
vCat is deployed at the edge, close to your tenant, and built as two products working together: an extension, which scans your Qlik Cloud Analytics estate and drafts descriptions for the master measures, dimensions and fields already in your apps; and a portal, where your data steward reviews each draft. Nothing is published to the estate as a governed description until a human has signed it off.
- Step 1Your Qlik analytics estateThe master measures, dimensions and fields already built into your apps.
- Step 2vCat extension and discoveryScans the estate and drafts descriptions, metrics and relationships.
- Step 3Candidate mappingDrafts become candidate business concepts: proposals, not facts.
- Step 4 · HumanSteward reviewYour data steward approves each one in the vCat portal. vCat proposes; your steward approves.
- Step 5Governed conceptThe trusted, authoritative business definition, written back to your apps.
That matters because Qlik Answers already prioritises master item descriptions over its own inference, when they exist. The gap in most tenants isn’t the prioritisation logic; it’s the descriptions themselves. Writing a clear, correct description for every master measure and dimension by hand doesn’t scale. vCat closes that gap with an automated, human-approved way to keep them complete.
In our own testing
Before governed descriptions were in place, Qlik Answers’ Thinking mode would occasionally fall back on the model’s training data rather than your data. After they were written back into the app, it stayed inside your data. That’s evidence, not a claim, and it’s the before and after we can show you against your own tenant.
The concept
A business concept is not a formula
A concept like Net Revenue is a governed signpost, not a calculation. Kept separate from the formula, one governed meaning can point to several implementations across your estate, without an agent ever guessing which one you meant.
REVENUEA governed business concept
The concept says what is meant; your Qlik apps compute it.
At runtime
Context first, implementation second
vCat and Qlik do two different jobs, in a fixed order. vCat MCP establishes the context before Qlik MCP computes the answer, so the agent understands what you mean before it chooses how to compute it.
- The question“Compare revenue against budget”Asked in plain language.
- OrchestrationRoutes the requestToday through how the skills are built; an orchestration layer is on the roadmap.
- vCat MCP · ContextIntent, meaning, metricSteps 1 to 3. Returns the definition, its trust level and how fresh it is.
- Qlik MCP · ImplementationThe calculationSteps 4 and 5. Runs the implementation and returns the number.
- The responseThe answer, with provenanceThe figure, the governed definition behind it, and its trust level.
Today that ordering is enforced by how the skills are built; a dedicated orchestration layer is on the roadmap. Either way, the principle holds: an AI agent should understand what you mean before it decides how to compute it.
Alongside Qlik
Where vCat sits alongside Qlik Answers and the Qlik MCP Server
vCat isn’t a replacement for either. It’s the layer that makes both work smarter with what you already have.
Qlik Answers
Qlik Answers already prioritises your master item definitions over model inference, and cites the fields and expressions behind every answer. That holds where a description exists. Where master items are undescribed, it falls back to unstructured content and general LLM reasoning. vCat keeps descriptions complete, approved and current, so that fallback is rarely needed.
The Qlik MCP Server
The execution layer: it gives an AI agent governed access to your apps, data products and datasets, runs the analytics engine and returns a computed answer, steps 4 and 5 of the path. vCat MCP sits ahead of it in the same request, resolving intent, meaning and the right metric before Qlik MCP computes anything.
VISEON Data Assurance
vCat tells your AI which number you mean. VISEON Data Assurance tells you whether that number can be trusted, scoring every row of the app data and QVD files behind it.
vCat is built by VISEON, not a Qlik product. Qlik, Qlik Answers and Qlik Cloud are trademarks of QlikTech International AB.
See it on your data
See vCat on your own tenant
We’ll show you the before and after against your own Qlik Cloud Analytics estate. VISEON for Qlik is from £12,000 per annum.