Skip to content
Module 02 · Web chat

Intelina DeskAI chat + ticketing tool

One script tag puts an AI support agent on your site. It recognises the signed-in user, answers from your live APIs and your knowledge base, and opens a ticket with a human the moment it is out of its depth — with the full agent, supervisor and admin workspace behind it.

Intelina Desk0:48chatting

Why was I charged twice this month?

I see two charges on your Pro plan. The second was a failed retry — it will drop off in 3 days.

I'd rather talk to someone.

Escalated · ticket #2041 assigned
Support agent

Thanks for waiting — I've refunded the duplicate charge. Nothing else you need to do.

endpoint · billing_historyescalated to agentticket #2041 · thread attached
Time to install
1 script tag
No SDK, no rebuild
Customer data stored
None at rest
Fetched live, cached 2 minutes
Handoff
Context intact
Agent sees the whole thread
Capabilities

Everything Desk does out of the box.

Intelina Desk replaces Static FAQ bots and shared inboxes. These are the parts that do the replacing.

Embed in one line

A single script tag, themed to your brand colour, logo and screen position. It renders on your site without touching your build.

Knows the signed-in user

The widget calls an endpoint on your own domain with the user's session cookie, so the bot greets them by name and answers personally from the first message.

Live data, not stale copies

Register your APIs as tools — wallet balance, orders, subscription plan. The AI decides when to call them and reads the answer back. Nothing is copied into our database.

You choose what the AI can see

Each endpoint carries a field whitelist, so only the exact fields you name are ever passed into the model. Everything else is dropped before it leaves your API.

Hardened outbound calls

Endpoint URLs must be HTTPS, private and reserved IP ranges are rejected, redirects are disabled, and the resolved address is re-validated at connection time to close DNS-rebinding gaps.

Knowledge base retrieval

Upload PDF, DOCX, TXT or Markdown into categories — billing, technical, onboarding, policies — and answers are grounded in the retrieved chunks instead of guessed.

Escalation that keeps context

When confidence drops or the customer asks for a person, the conversation becomes a ticket and lands with a human who can read every prior message.

A real ticketing workspace

Assign, reassign, close, reopen and rate tickets. Queues per team, SLA tracking, and a supervisor view across every agent.

Approvals before sensitive actions

Actions you flag as sensitive wait in an approval queue for a supervisor or admin instead of being taken by the AI on its own.

Real-time everywhere

New messages, assignments and notifications are pushed to agents and customers the moment they happen — nobody refreshes a page.

Analytics and knowledge gaps

Agent performance, resolution rates, project and cross-project reporting — plus the questions the AI could not answer, so you know what to document next.

Audit trail by default

Admin operations, data fetches, authentication events, escalations and key changes are all recorded with actor, action and entity — metadata only, never payloads.

Runtime

What happens when a customer opens the chat

  1. 01

    Widget loads

    The script authenticates with a project-scoped API key and asks your site who the current user is.

  2. 02

    Question arrives

    The message is checked against escalation rules before anything else runs.

  3. 03

    Context gathered

    Registered data endpoints are called for live values, and the knowledge base is searched for relevant passages.

  4. 04

    Answer generated

    The model answers from that context, calling further tools within the turn if it needs more.

  5. 05

    Escalate if needed

    Low confidence or an explicit request converts the conversation into a ticket and assigns a human.

  6. 06

    Resolve and learn

    The agent closes the ticket, the customer rates it, and unanswered questions surface as knowledge gaps.

Inside the dashboard

The screens your team actually works in.

Not a feature list — this is how the product is laid out once you log in.

Chat Widget

What your customers see.

  • Themed colours and logo
  • Position control
  • Signed-in user detection
  • Conversation history

AI Policy

The rules the bot plays by.

  • Escalation thresholds
  • Tone and behaviour
  • Response guardrails
  • Per-project overrides

Knowledge Base

Documents the AI can quote.

  • PDF / DOCX / TXT / MD upload
  • Category tagging
  • Chunk indexing
  • Retrieval search

Data Endpoints

Live APIs the AI can call.

  • Endpoint registration
  • Field whitelisting
  • Signed short-lived tokens
  • Test runner
  • Secret rotation

Tickets & Queues

Where humans take over.

  • Assignment and reassignment
  • Close, reopen, rate
  • Team queues
  • SLA tracking
  • Escalated chats

People & Roles

Who can do what.

  • Super admin
  • Project admin
  • Supervisor
  • Support agent
  • Two-factor authentication

Analytics

How support is actually doing.

  • Project analytics
  • Agent performance
  • Knowledge gaps
  • Reports and digests
  • Global view

Governance

Controls for regulated teams.

  • Approval queue
  • Audit log
  • API keys
  • Event webhooks
  • Per-project isolation
Good fit

Where it earns its keep.

Teams that get the most out of Intelina Desk usually look like this.

  • SaaS and marketplace products with logged-in users
  • Billing, wallet and order questions
  • Support teams tracking SLAs across queues
  • Any team replacing a shared support inbox
Also covering phone?

Intelina Voice handles ai voice agents + customer crm on the same platform layer — the knowledge you upload here is reusable there.

Explore Intelina Voice

Bring one support queue. We'll show you the rest.

A working Desk demo on your own knowledge base and your own systems — not a canned recording.