Intelina
AI that answers the phone and the chat window
Intelina automates customer conversations across two channels that usually live in separate tools. Voice answers inbound calls with real-time speech-to-speech agents; Desk is an embeddable chat widget with a full ticketing workspace. Both read from one knowledge layer and can act on your live systems mid-conversation.
- <60s
- Knowledge freshness
- 2
- Channels, one brain
- Live
- API grounding
- 3
- Role tiers
{ why it exists }
Answering is easy. Resolving is the hard part.
Support automation needs three things at once: knowing who is calling, knowing the current answer, and being able to do something about it. Most tools deliver one of the three, which is why customers still end up waiting for a human — and why deflection metrics look good while resolution does not.
- Callers repeating details the business already holds
- Chatbots answering from stale copies of the help centre
- Assistants that explain a policy but cannot act on it
- Escalation without context, so the customer starts over
{ what it does }
What Intelina does
Speech-to-speech voice agents
Real-time voice models with natural turn-taking and interruption recovery, rather than a chained transcribe-think-speak pipeline that stacks latency until the call stops feeling human.
Caller identification
The caller is recognised on connect and their account, order or case history is loaded before the agent speaks — nobody reads out a reference number they already gave the IVR.
Knowledge grounding
Upload PDFs, DOCX and TXT into a shared retrieval layer serving both channels, with new content reflected in answers in under a minute instead of after a nightly rebuild.
Custom tools via webhooks
Define authenticated actions the agent may take — check status, book, reschedule, raise a ticket — and it calls them mid-conversation against your real systems.
Embeddable chat & ticketing
One script tag installs Desk, which recognises authenticated sessions, reads live data through endpoint whitelists with field-level access control, and escalates into a ticket workspace.
Transcripts & call history
Agent, supervisor and admin roles over a workspace with recordings, transcripts and searchable history for quality review, coaching and audit.
How Intelina is built
Two channels, one intelligence layer. The engineering sits in the parts users never see — the latency budget, the retrieval freshness guarantee, and the boundary around what an agent is allowed to do.
Conversation Layer
Real-time speech models with turn-taking, barge-in handling and interruption recovery, tuned against a fixed end-to-end latency budget.
- Speech-to-speech models
- Turn-taking
- Barge-in handling
- Session state
Knowledge Layer
A shared retrieval service both channels query, with incremental indexing so a document change propagates in seconds rather than overnight.
- Document ingestion
- Chunking & embedding
- Hybrid retrieval
- Incremental sync
Action Layer
Tool definitions bound to authenticated HTTP webhooks, with endpoint whitelists and field-level access control deciding what the agent may read or change.
- Function calling
- HTTP webhooks
- Endpoint whitelists
- Field-level ACL
Operations Layer
Everything the support organisation needs around the agent: escalation rules, ticket workspace, recordings and searchable transcripts.
- Ticket escalation
- Role hierarchy
- Transcripts
- Call history
{ built for }
Who uses it
Healthcare providers
Appointment booking, rescheduling and refill requests handled without a queue.
Financial services
Account and transaction questions answered from live systems with field-level access control.
E-commerce & retail
Order tracking, returns and delivery questions resolved on first contact.
Insurance
Claims status, policy questions and document requests handled around the clock.
Telecommunications
Tier-one technical support and billing deflected before reaching an agent.
{ what it delivers }
Why it matters
Answers on first contact
Caller context and live system access mean common requests resolve without a transfer.
One knowledge base, both channels
Phone and chat answer identically because they read the same retrieval layer.
Actions, not just answers
Tool calling turns the agent from an explainer into something that completes the task.
Escalation without repetition
Humans receive the transcript and context, so the customer never starts over.
Intelina is live today.
Try it yourself — or tell us what you want built and we will scope it the same way we scoped this.
See all seven products