2024 Year in Review: The Year We Proved It Works.

When we started NoMind, the question we got most often was: can you actually build AI that works for Bhutanese institutions? Legacy systems, constrained infrastructure, a language that almost no AI model understands, and organizations that had never deployed anything like this before.

2024 was our answer.

By year end we had 7 AI agents in production, 100,980 customers served, and 144,401 queries resolved across institutions that Bhutan depends on every day. Users were coming from 9 countries. And we were only just getting started.

The 2024 Numbers

What We Built

Three institutions formed the foundation of our 2024 deployments: Bhutan Development Bank, Bhutan National Bank, and Drukair. Each one taught us something the previous one had not.

At Bhutan National Bank, the challenge was the infrastructure underneath. BNB runs on SOAP and XML systems with no modern API layer. We built middleware from scratch to connect a 2024 AI layer to legacy banking software without touching the core system. It worked on day one.

Bhutan Development Bank came with a wider customer base and more diverse query types. We had to build significantly better intent classification and escalation logic. Queries ranged from account questions to loan inquiries to branch location lookups. The agent handled all of them.

Drukair required domain-specific knowledge. Airline queries have tight logic: routes, baggage allowances, check-in windows, fare rules. Getting that context right without hallucinating incorrect information was the primary engineering challenge. We solved it.

Three deployments. Three different hard problems. All three in production by end of 2024, handling real customers every day.

What 1.43 Queries per Customer Tells Us

The average of 1.43 queries per customer in 2024 means most users came, asked one question, got an answer, and left. The systems were functional. They were not yet truly conversational.

A user who asks one question and leaves is a system that resolved a query. A user who asks five questions and returns the next week is a system that has become genuinely useful. We knew that gap needed to close, and we built the 2025 roadmap around closing it.

Where the Users Came From

Bhutan dominated by volume, as expected. What we did not expect was how strong the diaspora signal was. Australia came in second, Canada third, Singapore fourth. Bhutanese communities abroad were using our agents to access information about institutions back home.

That was unplanned, organic, cross-border reach. It told us the market was larger than just Bhutan's domestic population.

What 2024 Proved

It proved that AI built specifically for Bhutanese institutions, by people who understand those institutions, works. It integrates with legacy infrastructure. It handles real users at scale. It earns institutional trust.

That was the hypothesis when we started. By December 2024, it was a demonstrated fact.

The 2025 question was not whether the model worked. It was how far we could push it.

Read our 2025 report to see what happened next.

Explore our latest solutions. /solutions