Enterprise AI CallcenterOn-premises
On-premises AI call centre for banks and financial institutions
Financial institutions cannot send customers' voice recordings to cloud services without a long audit. AI Fortis runs on your server: speech recognition, voices and recordings stay inside your network.
What it solves
- Strict rules on moving personal data outside the institution.
- Peak load (due dates, paydays, incidents).
- Expensive 24/7 cover.
How it works
- Hotline: identification, balance, card status, fees, opening hours.
- Outbound: instalment reminders before and after the due date, with promise to pay.
- Surveys after a branch visit or a call.
- Transfer to staff with a summary for complaints or complex cases.
Example call
Illustration with invented names and data. Script, tone and rules are yours. The agent speaks Bulgarian; the lines are translated here.
Integrations
- Core banking / CRM through HTTP tools inside your network
- Your PBX over SIP
- Webhook to DWH or ticketing
Compliance
- Speech recognition and voice synthesis on your server; the language model is your choice (cloud API under a data processing agreement, or a local model, in development).
- The agent says at the start that it is an AI assistant (EU AI Act, Art. 50) and announces that the call is recorded.
- Audit log of platform actions, roles and users per client.
- Recordings and transcripts stay on your server; retention is set per client.
What we measure in the pilot
- Share resolved without staff
- Time to answer at peak
- Promises to pay
- Post-call NPS
We will publish results from real deployments with the customers' permission.
Frequently asked questions
Does everything really run on our servers?
Yes: our own speech recognition and voice models, conversation control, the telephony gateway, campaigns and recordings. The language model is your choice: your existing provider via API (then only the text of the turns is sent) or a fully local model with no internet.
How fast does the agent answer?
In our test calls with a cloud language model the median time from the end of the caller's sentence to the first sound of the answer was 1.4 s; the voice itself starts 136 ms after the answer text is ready. The language model is the largest part, so we measure it with your choice in the pilot.
Which telephony does it work with?
Any standard SIP trunk or PBX: your existing numbers and carrier stay in place.
Does the agent say it is an AI?
Yes. At the start of the call the agent introduces itself as an AI assistant and announces the recording (EU AI Act, Art. 50). The wording is configurable.
Get a free sample agent for "Banking and finance"
Tell us your call type and language; we build a sample agent you can call and judge yourself.
- GDPR by design
- Runs on your own servers
- 24/7 operation and support
- A team of 5 in Sofia, EU
Already running in production at organisations in Bulgaria — names under NDA.