Industry · Banks in Armenia
AI assistants for banks in Armenia
tigoLab is a Yerevan-based team that builds AI assistants for banks: customer replies in Armenian, Russian, and English, loan document intake, and an internal knowledge bot for staff. Each one runs in your infrastructure, and a person approves anything customer-facing.
Built by Narek Maloyan and Tigran, a founder-led team in Yerevan, Armenia. Narek built the agent that reached #1 on SWE-bench Verified (70%, May 2025), and has 7+ years of production machine learning and a PhD. You talk to the engineers who build your project.
Where an assistant helps a bank
Staff knowledge bot
Branch and call-center staff ask about products, tariffs, and internal procedures, and get an answer from your own documents with the source linked. Built and measured like our RAG knowledge bots with evals.
Customer replies
The assistant drafts replies to routine questions from email, chat, and messaging channels, in the customer's language. A person approves before anything is sent.
Loan document intake
It checks that an application pack is complete, reads the key fields, and flags gaps or mismatches for the loan officer. See the document checker demo for the approach.
Back-office workflows
If you run on Microsoft 365, we build Copilot Studio agents and Power Automate flows in your tenant. See Microsoft 365 automation.
How we work with a bank
- We start with one internal use case, where the risk is low and the time saved is easy to count.
- We build in a test environment with masked or synthetic data, never on live customer data first.
- The assistant runs in your infrastructure or cloud tenant, with the model endpoint your IT team approves.
- We agree the acceptance tests in writing and measure them, so the decision to go live rests on numbers.
- A person approves every customer-facing message. Your rules decide where the approval sits.
Three languages, tested
Customers in Armenia write in Armenian, Russian, and English, and often mix them. We build the test set in all three languages with your team, and we report the quality per language. If one language is not good enough for a task, the assistant hands that case to a person.
Questions
Which languages does the assistant work in?
Armenian, Russian, and English. We test the answer quality in each language with your team before anything goes live.
Where does our data stay?
Inside your infrastructure or your cloud tenant, with the model endpoint your IT and compliance teams approve. We do not train models on your data, and we keep no copy on our side.
How do we start without a long procurement?
With one internal pilot, such as a knowledge bot for branch staff, in a test environment with masked or synthetic data. It has a fixed scope, a fixed price, and written acceptance tests.
Are you based in Armenia?
Yes. tigoLab is based in Yerevan, so we can meet your team in person.
Related: data security, knowledge bots with evals, and about the team.
Start with one internal pilot
Book a call, or meet us in Yerevan. We will propose one use case, the acceptance tests, and a fixed price.