VNClagoon AI

VNClagoon AI is the AI layer of the suite. It is not a chatbot bolted onto a product: the same assistant, the same grounding and the same policy apply in every application. The design goal is that an organization can use AI on its own internal content — mail, chats, channels, files, projects, customers — without that content leaving its control.

Three decisions define it:

1. Confidential by default. Requests follow the organization’s policy. The models can run inside the organization’s own environment, and content classified as confidential or restricted never leaves the perimeter.
2. Grounded, not guessing. Answers come from the organization’s own content — the channels, files, records and conversations the user may access — and they cite their sources.
3. An agent that acts carefully. Where the assistant can change something, it proposes the action and executes it only after confirmation.

What it delivers

• One assistant in every product. Ask about the app you are in, or about a specific channel, file set, deal or request.
• Answers with sources. The user can see which document or topic the answer came from — which is what makes an AI answer usable at work.
• Semantic search. Find things by meaning, not only by exact words.
• Work reduction. Summaries, minutes, drafts, enrichment and next actions across the suite.
• A policy the organization controls. Local, private or an approved external model — decided by the organization, with confidential data never sent out.
• Auditability. Model provenance, budgets and an audit trail around AI actions.

Main capabilities

Capability	What it gives you
Assistant (Lago)	A consistent assistant in every product, grounded in the current screen and its context.
Grounded answers	Answers from channels, topics, files and records, with cited sources.
Semantic search	Meaning-based search across the suite’s content.
Summaries and minutes	Summarize a conversation, a topic or a meeting; produce minutes.
Drafting	Draft and refine mail, topics and records under policy.
Enrichment	Summarize a contact or company, suggest tags and propose a next action.
Next actions and risk	In VNCcrm, propose the next step per deal and score deal risk.
Ask your files	Read-only answers over the files a user may access, with citations.
Confirm-then-execute	Write actions are proposed and only carried out after the user confirms.
Local option	Run a request on the user’s own computer where a surface offers it, so nothing is sent anywhere.

AI policy and data handling

An organization chooses how AI is provided:

Policy	Meaning
On your infrastructure	Models run inside your own environment. Nothing leaves the perimeter.
Approved external model	For a defined use case, an approved external model may be used.
Local on the device	On surfaces that offer it, the user runs the request on their own computer.

Independently of the policy, content classified as confidential or restricted is not sent to an external model. Where an external model is used for lower classifications, the request is abstracted and screened first, and the response is validated before it is used — the model is never the system of record.

Typical use

• An employee asks the assistant in VNCchannels what was decided about a project and gets an answer with the two topics it came from.
• A team member asks Ask your files what the travel policy says about receipts and gets the cited paragraph.
• A sales representative sees a next action per deal with the reason behind it and confirms it in one click.
• An organization runs the assistant entirely on its own infrastructure, with confidential material never leaving the perimeter.

How it fits the suite

• Every product embeds the same assistant, so behaviour and policy are consistent.
• VNCdirectory supplies identity, classification and policy.
• VNCfiles, VNCchannels, VNCmail, VNCcrm, VNCproject, VNCtask and VNCworkflow provide the grounding content.
• Confidential Computing protects data while a model processes it.

Roles

Role	What they do
End user	Ask, summarize, search, draft and confirm actions.
Administrator	Set the AI policy, models, budgets and the allowed external providers.
Security / compliance	Review provenance, audit and data-classification rules.

Related

• AI options in VNClagoon
• Use the assistant
• Confidential Computing — protecting data during processing