Every call answered in Greek, every call summarised, on infrastructure you control.
A Greek voice agent for inbound service, IVR and call summarisation, and a RAG helpdesk for your agents. Deployed on your premises or on dedicated capacity in Greece, with an AI Act (Regulation (EU) 2024/1689) dossier for every model.
At a glance
- Deployment
- On-premises, your own cloud (BYOC) or dedicated capacity in Greece
- Languages
- Greek first, plus English
- Models
- Open-weight only, fine-tuned on Greek telephony speech
- Compliance
- AI Act dossier, model card, drift monitoring
- Pricing
- Per seat and per outcome, never metered per token
Our inference stack runs in production on contact-centre workloads for a Greek telecom operator and a Greek bank; the AI Employee platform is live at a leading Greek publisher. See the measurements
Three workloads, one governed stack
Greek voice agent
Handles inbound service in natural Greek: caller identification, FAQs, order and case status. Routes the call to an agent with a structured summary, so nobody asks the customer twice.
IVR and call summarisation
Every call summarised in Greek seconds after hang-up, in your CRM's format. Measured at 200 concurrent summaries per GPU within a 2-second KPI.
RAG helpdesk
Answers for agents grounded in your policies and procedures, with the source document attached and faithfulness scored on every release.
On-premises or BYOC
Runs in your data centre, in your cloud tenancy, or on dedicated capacity in Greece. No US API sits in the serving path.
AI Act dossier
Each model ships with a model card, a training-data manifest, evaluation results and a drift-monitoring plan your compliance team can file.
Guarded actions
Designed so that no action runs on the model's word alone. Every action passes through typed tools and rule checks and is written to an immutable audit trail.
From the first call to production
Measure on your calls
We replay a sample of your recorded calls and documents through the stack and report quality and latency before you commit.
Go live on one queue
One line of business, one queue, with hand-off to an agent and a daily quality review.
Scale and file
Move to production on dedicated capacity, with the AI Act dossier, monitoring and incident log in place.
The speech stack
- Greek speech recognition: Clio, tuned on Greek telephone audio
- Agent model: Metis 26B with typed tools and a guard layer
- Greek voice: Calliope, first audio in 0.19 seconds
- Text tasks: Iris 8B for call summaries; Metis 12B for document Q&A
Priced on outcomes
Pricing
Seats, a platform fee and a compliance pack, priced against the outcome: handled calls and summaries. Never metered per token. Volumes and the deployment model set the quote.
Compliance pack
Evaluations on every release, drift monitoring, an incident log and the AI Act technical file, maintained by us.
Start with one workload.
Tell us the calls, documents or questions you want handled. We measure it on your data before you commit.