A returning migrant in eastern Nepal is trying to decide whether to leave again and how to tell his family. He has questions — the kind you would normally bring to a mentor, if you had one. He doesn’t. He opens iLeader on his phone. He asks. His coach remembers their conversation from last week. It builds on it. It pushes back gently when his thinking is fuzzy. It helps him work out what to do next. 

That conversation, in Nepali, at his pace, on his phone, did not exist a year ago.

Migration is older than us

Migration itself, though, is not new. It is older than our species. Long before the first anatomically modern humans existed in Africa 300,000 years ago, our ancestors had been walking out of the continent for nearly two million years. Every human alive today is here because someone before them, somewhere, decided to leave. People will keep moving — for economic, social or environmental reasons.

The question has never been whether people will move. It has been whether, when they do, they have what they need to choose where to go, to settle into a new place, and — if they want to — to return home even better off than when they left.

The gap that started this

In 2023, remittances to low- and middle-income countries reached $656 billion — more than foreign direct investment and official development assistance combined. In Nepal, over 2,000 people leave for work abroad every day. Their remittances are equivalent to roughly a quarter of the country’s GDP. For many families, that money is the difference between rent paid and rent owed. School continued and school dropped.

But for the migrants, particularly temporary migrant workers, the journey is rarely linear. They go. They work for a few years. They come home to a place, a family and a community that has changed. And nearly 40% of them start the cycle again. At each stage, decisions made in moments of pressure shape what comes next. A loan taken to finance the departure. A contract signed in a language no one explained. A return that comes earlier than planned. A reintegration that doesn’t hold.

Helvetas’ research in Nepal identified 12 key moments in this journey. These moments carry significant emotional and economic weight. Too often, migrants face these high-stakes turning points completely alone, relying on fragmented advice from friends that can inadvertently lead to misinformation. Ultimately, how a migrant navigates these pivotal moments can make or break their entire journey.

Helvetas Nepal has spent over a decade walking alongside these migrant workers through two distinct projects: the Safer Migration project introduced in 2011 and the Reintegration of Returnee Migrant Workers (ReMi) project initiated in 2022.  As bilateral initiatives of the governments of Switzerland and Nepal, together with Ministry of Labour, Employment and Social Security, they reach over 500,000 people every year — through Migration Resource Centers before departure and Employment Service Centers on return. But this infrastructure is concentrated where it is most accessible. The migrants who most need support are often the ones furthest from it: in rural districts, in informal sectors, and in destination countries where no Nepali institution can reach them at all.

So the questions were raised: Could AI coaching extend support to where in-person services cannot go? In the migrant’s own language, at their own pace, and on their own phone? Can AI truly adapt to the vastly different realities of each returnee — whether they are coming from the Gulf or Europe, possibly carrying a deeply negative experience? And crucially, can it provide the much-needed privacy they require to ask the sensitive questions they would never dare ask someone face-to-face, all from the safety and comfort of their own homes? All this, without losing what makes coaching in the first place? 

Enter iLeader. Developed by Besage.ai, this AI-powered coaching tool is currently being piloted by Helvetas Nepal to extend personalized support to underserved areas. Through an interactive voice or text chatbot, the app combines individualized mentoring with realistic role-play scenarios to train returnees in critical skills.

Nepal, on purpose

Besage chose Nepal because it was hard. It also has one of Helvetas’ strongest programs and teams. If a coaching app could work here, it could probably work in most places we want to go next. So we volunteered for the deep end of the pool. What could go wrong? Right?

Oh, the learnings came fast and furious! … Customization meant rebuilding the coaching conversation in Nepali. Not translating it — rebuilding it, then testing it with native speakers, then rebuilding parts again because the first version sounded too formal, and again because the second wasn’t fluent enough in the everyday Nepali our Coachees actually use. It is working now. It didn’t three months ago.

It meant adapting coaching frameworks that came out of Western individualism — “set a personal goal and pursue it” — into frames that land more naturally with people who understand themselves through duty to family and community. We added a counselor-style mode so people can ask direct questions about ReMi’s services and get useful answers before the more reflective coaching begins. Coaching takes practice in introspection, and not everyone has practiced introspection. So we walk people through it.

There were “simpler” tasks too: It meant rebuilding the login system because most of our Coachees don’t have email. And unexpected ones(!): What happens if a Coachee cannot enter their date of birth or a goal’s deadline because for the system that date is 20 years in the future? Because you know, in Nepal, right now, the year is 2082! (Bikram Sambat calendar). Then, the app is already wrong.

That one last detail is what we tell people when they ask what “customization” means.

Besage is paying for this work itself — over $100,000 in engineering time and resources. Helvetas Nepal is covering its own costs in parallel, with on-the-ground teams working above and beyond their scope of work. We do it because we believe it must be done, in a difficult context, before anyone could honestly claim AI coaching can serve emerging economies well.

The partnership behind this work has a clean shape: Besage owns the technology. Helvetas Nepal owns the operational know-how — over a decade of trust with migrants, counselors and the government that no software can replace. Together, we co-own the learning.

What are we seeing, so far 

It is early days and yet lessons are reaching us loud and clear. The first one: You cannot introduce a digital human-development tool in rural communities by virtual communications (such as email/WhatsApp) only.

Enthusiasm among migrant returnees has been consistently high. But digital literacy varies. Connectivity is uneven. And trust in digital tools is fragile — many of these participants have been scammed online during their migration.

We knew this. It is one reason iLeader is deliberately built as a B2B product: It lives inside a partner’s service portfolio, alongside in-person counseling and program staff, not simply as something a migrant downloads on their own. Even so, what we thought would be a digital journey turned out to need a great deal of in-person work during the onboarding phase. In Nepal, municipal Focal Persons and Counselors led outreach. A part-time consultant was brought in to augment them. Together, they became the bridge that made the whole thing work. The technology is necessary. It is never sufficient. Trust cannot be engineered through a phone; it requires facilitation from someone they already know.

And then, for the people who broke through that initial friction, something remarkable started happening: Field reports describe migrants using the coach as a private space to process the anxiety of return, manage stress, and navigate complex emotions. iLeader also proved its adaptive worth: One returnee used it as a business mentor, retaining context from previous conversations and giving him deeply tailored advice. This is a use case on the economic front we didn’t design for, yet possibly as important as the use cases we did.

“[It] teaches us how to express our inner thoughts and feelings. Things we often cannot share or discuss with our friends.  Seeking guidance from [the coach] really helps understand and process them much better.”

These are early signals. Not outcomes. The first meaningful data on whether soft skills actually moved and impact was achieved will come over the next few months. We will share what we learn, including the parts that don’t work.

What we don’t know yet

 A lot. Honestly.

We don’t yet know whether AI coaching reaches the migrants who most need it, or mostly the ones who were already best positioned to benefit. We don’t know whether coaching earlier in the cycle — before departure — changes outcomes more than coaching only at return, or at any other single point among the 12 moments that matter. We don’t know how much coaching it takes before a person has a real breakthrough. Two hours? Ten? Thirty? We don’t know whether voice is worth the higher cost compared to text.

And there is also the question about what “sustainable” means for an AI-powered intervention in a context where governments are stretched and humanitarian funding is shrinking. The promise of AI coaching is not that it is free. It is that it is dramatically cheaper than a one-to-one human coach, available anywhere, at any hour, at scale, in any language.

The cost structure of AI coaching — high up front, low per-conversation, with a recurring tail — is not unique. It is the cost structure of nearly every technology that has ever been brought to underserved populations: vaccines, improved seeds, mobile money, education itself. Each one needed external (donor) subsidies for years before reaching the point where governments and beneficiaries could share the cost. AI-services such as AI coaching are at year one of a similar arc.

This pilot is fundamentally about learning, and Helvetas Nepal, with Besage’s support, is looking at several new avenues to build a deeper understanding of how these tools perform over time. One such opportunity includes a collaborative research initiative designed to rigorously unpack these complex dynamics. In partnership with the University of Zurich and supported by Besage, a multi-institutional research proposal has been developed to study these empirical questions systematically across the full migration cycle, bringing together academic expertise from Zurich, Michigan, Amherst, and Kathmandu. Whether through this specific grant or other exploratory avenues, we will keep learning. Our analytical system is built for learning and questioning: what works, what doesn’t, what explains it.

The deeper point is this: Coaching has been studied for decades. There is strong evidence — from meta-analyses across hundreds of studies — that real coaching helps people make better decisions, build confidence, and follow through on commitments they make to themselves. That evidence is the foundation we are building on. What is genuinely new — and unproven — is whether AI coaching at scale can have a similar impact, and for people whose lives look very different from the populations the research was done with.

Change is happening all around us

AI is moving fast. The platform that is cutting-edge today will be legacy tech soon enough. We are designing for that, deliberately.

The foundation is now laid. The pilot has established that while the capacity for digital transformation is immense, bridging the digital divide requires a human scaffold. The future of this work hinges on a dual commitment: the technological innovation and agile engineering of Besage, operating in perfect synergy with the deep-rooted community outreach and human-centric infrastructure of Helvetas and the ReMi project.

Our intention is that a Coachee uses iLeader for the time they need it — six months, twelve months, however long it takes for the durable capacity to build — and then they move on. The technology is a catalyst. The lasting change is in the person and the community: the digital fluency, the agile mindset, the confidence to navigate the next decision, and eventually, build a life better than the one they left.

Our commitment is to the migrant: the technological innovation and agile engineering of Besage to keep evolving iLeader and build other accompanying tools, operating in perfect synergy with continued deep-rooted community outreach and human-centric infrastructure of Helvetas Nepal. What matters is that quality coaching, together with improved tools and infrastructure, customized to the people it serves, with outcomes that can be measured, becomes possible at scale — for the migrants doing the work that holds up global supply chains, for the students whose universities don’t have career services budgets, for the community health workers in places no telemedicine company will ever serve.

Most AI tools in the world today are built for higher-income consumer markets. English-speaking. High-tech. With guaranteed email addresses and stable Wi-Fi. What we are testing in Nepal is whether AI can be built deliberately for everyone else — for the people the technology wasn’t built for.

We have many questions.  A handful of hard-earned answers. And a relentless drive to keep learning.

About the Authors

Elizabeth Gozzer is the CEO and Co-Founder of Besage.ai, an AI-native social enterprise at the forefront of applying artificial intelligence for human development.

Madushika Lansaka is the Team Leader for the Reintegration of Returnee Migrant Workers project at Helvetas Nepal.