| Result | in-house LLMs solve 45% of requests; support costs estimated down at least 25% a year |
| Source | Digital Platforms, case no. 2447 |
| A similar task for us | Customer support |
| Industry | Banks and finance |
What was done and what came out
- The copilot reads messages and listens to calls in real time, suggests an answer and the next step, and instantly finds the reason, the amount and the date of a disputed transaction.
- Customer issues are solved 10% faster.
- The bank estimates that its own models cut support costs by at least 25% a year; it recorded direct savings during six months of testing but did not disclose the amounts.
What to take for yourself
An in-house model pays off where requests come in by the thousands every day: the bank keeps both the model and the data inside and gains on every single request.
AI agents on a local model in your network: a pilot on one process. A pilot AI agent on a local model in 14 days from $1,499 with the metric in the contract, on your hardware or on rented GPUs. An AI department of 3–5 agents in 4–6 weeks from $4,999. You buy the hardware to our specification.
Sources and fact-checking
The facts and numbers on this page come from the publication. Data type: project catalogue entry. The company is not our client: the page shows what a smaller company can repeat from this experience. Checked by the editors on 30 September 2026.
Found a mistake? Tell us
Similar cases
- Online store customer support: common questions without operators
- Marketplace review replies: an answer to every review and question within an hour
- A large Russian financial group: AI financial assistant on in-house language models, beta since 20 July 2026
- Alfa-Bank, IT head: local agent on the Qwen 3.6–35B-A3B model: about 70 tokens per second on a Mac
- T2 (Tele2): tender for a self-hosted AI coding assistant in a closed data center: 15 developers, response within 400 ms
- All materials in the industry: Banking
Prices and measurements in this section were checked against the sources at the bottom of each page. We build AI agents on cloud and local models and do not sell hardware. Calculations are estimates; legal conclusions on data protection come from your lawyer.