Integrations
Language models
The language model providers Telonic works with, how a model is chosen for your deployment, and how that choice stays within your region.
On this page
The language model is the part of the agent that understands the conversation and composes the reply. Telonic works with models from several established providers and chooses the right one with you for each deployment, based on your languages, your use and your region. The model decides how an answer is expressed.
The facts in the answer come from your systems and approved documents, and the actions available to it are the ones your team has granted. All processing for your deployment runs in the region you choose, unless you explicitly choose a language model provider outside it.
What the language model does in your deployment
| Role | What it means for you | Example |
|---|---|---|
| Understand the request | Works out what the customer wants, including in Gulf, Egyptian and Levantine Arabic, and when they switch between Arabic and English | "أبغى أغير الحجز to Friday" understood as a change of arrival date |
| Compose the reply | Expresses facts retrieved from your systems in your tone, in the customer's language | A booking change confirmed in the tone you set |
| Choose the next step | Selects from the actions your team has granted, within your limits | Offering a person when a request is outside the agent's limits |
| Read images and documents | Reads what customers send, using a model running in your chosen region | A photographed police report attached to a claim |
Providers Telonic works with
| Integration | What the connection covers |
|---|---|
| OpenAI | Language models |
| Anthropic | Language models |
| Google Gemini | Language models |
| Mistral AI | Language models, including open-weight models |
| Meta (Llama models) | Open-weight language models |
| TII (Falcon models) | Open-weight language models |
Which providers and models are available in your region is confirmed during implementation, and named in your agreement. On-premises deployments run open-weight language models (models whose weights are published, so they can run on your own infrastructure), with computing capacity sized with you during implementation.
How the choice follows your region
All processing for your deployment runs in the region you choose, unless you explicitly choose a language model provider outside it. The language model is the only component that can be chosen outside your region. Speech recognition and voice synthesis always run in your chosen region.
If you choose an out-of-region model, personal information is redacted (removed or masked) before any text leaves your deployment. It is replaced with placeholders, and the real values are restored in the reply inside your deployment. Redaction covers names in Arabic and Latin script, phone numbers, email addresses, national identity numbers including Emirates ID and Iqama numbers, passport numbers, payment card numbers, bank account numbers and IBANs (international bank account numbers), dates of birth and street addresses. See Personal information redaction.
How providers handle your data
Telonic's contracts with every provider prohibit using your data to train or improve their models. Zero data retention arrangements (where the provider keeps no copy of what it processes) are used wherever the provider offers them. Every provider used for your deployment is listed in the sub-processors (companies that process personal data on Telonic's behalf) for your deployment. See Sub-processors.
The technical detail
Models are reached through a model routing layer (software that sends each request to the provider chosen for that role), so a model can be changed without rebuilding the agent. Your industry model, limits, memory and integrations stay the same when it does. A new model is adopted only when it performs at least as well on the industry test set, and any change of model or provider needs your approval before release.
If a provider has an outage, failover only ever switches to an approved provider in your chosen region, and fallback providers are tested in advance. Calls travel over TLS (Transport Layer Security), version 1.2 or higher.
Permissions typical for this category
Every action in your systems is taken by Telonic's software, under the permissions your team has granted. The language model receives only what the agent sends for each turn of the conversation, and has no connection of its own to your systems. See Permissions.
In practice
Nakhla Travel, a tour operator, sets up its deployment in the UAE region.
- During implementation, Telonic confirms which language models are available in the UAE region and tests them on the operator's own itinerary conversations.
- The operator keeps every provider in-region, and the chosen model is named in its agreement.
- Months later, a newer model becomes available in the region. Telonic tests it on the same conversations and it performs at least as well.
- The operations director approves the change, and it goes live.
What your team controls
- Whether any language model outside your region may be used.
- Approval of every change of model or provider before it goes live.
Related
- Models and providersAgents
- Hosting and data residencySecurity and data protection
- Personal information redactionSecurity and data protection
- Sub-processorsSecurity and data protection
- Testing before every releaseGovernance and control
Product names and logos are trademarks of their owners. Their mention shows systems Telonic connects to and does not imply partnership or endorsement.