AI for banks and financial institutions
Banks and lenders use AI Fortis for three things: AI voice agents that answer the hotline and make payment reminder calls, document capture that turns paper into checked data, and integrations with core systems. The Enterprise AI Callcenter runs on the bank's own servers, so call audio and customer data never leave its premises.
Where AI takes work off a bank's teams
Customer hotline
Balance and card questions, branch hours, how-to questions and the status of a request, answered at any hour. The agent says it is an AI and passes the call to a person with what was said.
Payment reminders and collections
Outbound calls before and after a due date, within the hours and attempt limits you set, never to a number on your do-not-call list. Promises to pay are recorded as structured data.
Lead and application follow-up
Calls back to people who started an application, to complete missing details or book a meeting with an adviser.
Documents and invoices
ScanSystems captures supplier invoices and other documents, extracts the fields, shows doubtful values to a person next to the original and archives everything per branch.
The rules that shape AI in banking
Since 2 August 2026, a voice agent must tell callers that they are speaking with an AI, at the latest at the first interaction (EU AI Act Art. 50). Calls and recordings are personal data, so the GDPR applies in full: a legal basis, information at the start of the call, security of processing and, for new technology at scale, a data protection impact assessment.
AI systems that evaluate the creditworthiness of natural persons or establish their credit score are high-risk under the AI Act (Annex III, point 5(b)), with the high-risk duties applying from 2 December 2027 after the Digital Omnibus on AI. A voice agent that answers questions, reminds people of payments and records what they said does not make that decision, and it should be set up so that it never does.
Under DORA, a bank's ICT third-party providers are part of its own risk management. Running voice AI on the bank's own servers keeps the call path out of an external cloud, which simplifies the third-party register and the exit plan.
Sources: EU AI Act (Regulation (EU) 2024/1689), Art. 50(1); EU AI Act (Regulation (EU) 2024/1689), Art. 50(5); EU AI Act (Regulation (EU) 2024/1689), Annex III, point 5(b); Digital Omnibus on AI (Regulation (EU) 2026/1744); GDPR (Regulation (EU) 2016/679), Art. 35; GDPR (Regulation (EU) 2016/679), Art. 32; DORA (Regulation (EU) 2022/2554), Art. 28
On-premises or cloud for a bank
| Question | Enterprise AI Callcenter (on-premises) | Web-based AI Callcenter (cloud) |
|---|---|---|
| Where call audio and recordings are processed | Inside the bank's own data centre | On our platform, with AI Fortis as processor under a data processing agreement |
| Effect on the ICT third-party register | No external processor in the call path | AI Fortis is an ICT third-party provider |
| Cost model | One-time from €150,000 plus €5,000 a month | €0.10 to €0.21 a minute plus integration |
| Time to first calls | Hardware, installation and integration first | Faster: no hardware |
| Suits | High volumes and strict data rules | Pilots, smaller volumes, quick start |
How a pilot in a bank runs
- Pick one scenario with clear volumes, such as reminder calls three days before a due date.
- Agree what the agent may say, what it must never say and when it hands over to a person.
- Connect the scenario to the core or collections system through the agreed interface.
- Test with internal callers, then a limited customer group, with every call recorded and transcribed.
- Measure answer rate, completed tasks, hand-overs and complaints against the agreed targets before scaling up.
Frequently asked questions
Can a bank run AI voice agents on its own servers?
Yes. The Enterprise AI Callcenter is installed on the bank's servers; call audio, recordings and transcripts stay inside its premises.
Is an AI voice agent a high-risk AI system for a bank?
Not by itself. Under the EU AI Act, systems that assess creditworthiness or set credit scores of natural persons are high-risk (Annex III, point 5(b)). A voice agent that answers questions and makes reminder calls does not take that decision.
Does DORA apply to AI voice agents?
DORA applies to the bank's ICT risk management, including its ICT third-party providers. An on-premises installation keeps the call path inside the bank; a cloud service is managed as a third-party arrangement.
Can the AI make collection calls?
Yes. It calls within your hours and attempt limits, respects your do-not-call list, says it is an AI and records promises to pay as structured data.
What does it cost?
The Enterprise AI Callcenter starts at €150,000 one-time plus €5,000 a month; the Web-based AI Callcenter costs €0.10 to €0.21 a minute plus a one-time integration fee.
Talk to us about your case
Tell us which calls or documents you want to automate first and where your data must stay. We reply personally, usually within one business day.
Pick a time for your call
Choose a slot that suits you. You get a confirmation with a video link.
30 minutesMon – Fri 09:00–17:00 (Eastern European Time)Google Meet
Open the calendar in a new tab →
Prefer not to pick a calendar slot? Tell us your preferred time instead.
Prefer the phone? Call +359 888 899 932. Or write to office@aifortis.com.