From International Economics to AI Product Lead: How Elfin Made Every Detour Count

I'm Elfin Li, I'm 26, and I'm based in the UK. I usually tell people I live at the messy, fascinating intersection of AI, product design, and international business. My trajectory has not been linear in the slightest. I started out in China studying International Economics and Trade, took a detour through sports tech and the energy sector, and eventually came to the UK to study Computer Science at the University of Bath.

If there is a common thread running through all of it, it is this: I thrive on building things from absolutely zero. Put me in an ambiguous environment without a playbook, and I am in my element. Today I work as a Product Lead focused on AI-enabled workflow transformation. In plain terms, I help businesses stop treating AI like a shiny new toy and start using it as a core operational engine.

I studied at Shanghai International Studies University, widely recognised as China's top institution for diplomats. My ambition at the time was exactly that: to stand on the world stage and tell China's story to a global audience. What happened instead was that I tumbled into product management, which turned out to be an entirely different form of diplomacy.

I spent years in high-stakes, large-scale environments: international sports governance, including the Olympic Games, and the new energy sector, sitting between senior board members, international federations, and technical teams on the ground. There is something profoundly moving about watching athletes lift their trophies knowing the system you built supported that moment.

Equally, seeing energy projects bring real benefits to thousands of families completely shifted my perspective. I realised building products is just another medium for connection and impact. That realisation is what drove me to the UK to formally study Computer Science, and eventually to AI.

When you operate at that scale, you very quickly hit the ceiling of traditional software and human bandwidth. I was in the operational trenches, so I saw exactly where the friction was. I saw where critical decisions stalled, where brilliant people were bogged down by repetitive administrative work, and where massive systems became dangerously brittle.

That was the situation that pushed me toward AI. Not because it was a fascinating new technology, but because it was the obvious, necessary solution to the bottlenecks I was watching destroy productivity every single day.

The crucible of the Olympic judging system was a particular catalyst. When you build software for an Olympic arena, it has to be bulletproof, entirely explainable, and trusted implicitly by the users. Taking that rigorous, high-stakes mindset into the early AI space was a challenge I simply could not resist.

My Computer Science master's at Bath gave me the academic grounding in data and systems, but you cannot learn AI purely from a textbook. My real education was entirely hands-on. I dug into documentation, used tools like GitHub Copilot and Claude daily, and forced myself to build.

Instead of just watching a video about RAG, I built a system. Instead of just reading about privacy, I figured out how to run models locally using Ollama. If I were starting again today, I would stop hoarding tutorials and start building on day one. Pick one deeply boring, painful business problem and just build a solution for it.

I would also document the journey publicly much earlier. Sharing your messy work-in-progress is the best way to build credibility, and it forces you to articulate what you are actually learning. The space moves too fast to wait until everything feels polished.

By flat-out refusing to sell AI. I simply looked at teams, identified where brilliant people were doing soul-crushing manual work, and offered them a system to fix it. The transition was gradual rather than a dramatic switch. Because I already had a background in product and stakeholder management, I started quietly injecting AI into the workflows around me without waiting for permission.

The formal opportunities evolved from the results I was already delivering. If you frame your work entirely around a business outcome, saving time or improving decision-making, people take you seriously very quickly. Say: I can build a system that saves you five hours a week on client proposals. Do not say: I know how to write prompts. Nobody actually wants to buy AI. They want to buy their time back.

It is highly practical and no two days are the same. Some days I am deep in product discovery, using models to process chaotic user feedback and turn it into structured requirements.

Other days are technical: stress-testing prompts, experimenting with local models, or figuring out how to integrate an AI tool into a legacy enterprise system. A typical example is stripping down a manual content pipeline and rebuilding it. I design a workflow that takes an input brief, enforces brand guidelines, generates a draft via AI, mandates human review, runs a quality checklist, and logs performance feedback.

I call this Intent Architecture. It is about defining the real user problem, pinpointing exactly where AI adds value and where it does not, and fiercely protecting what I call the Human Alpha: the strategic human judgement that keeps the output commercially viable and reliable, not just artificially generated.

Obsessing over the tool instead of the problem. People constantly ask me which AI they should learn. It is the wrong question. Ask yourself what specific operational bottleneck you are trying to solve. If you just chase the shiniest new demo without a commercial use case, you end up with a flashy toy that produces zero actual value.

There were so many new models, tools, and papers dropping daily that I felt I had to master all of them simultaneously. It was exhausting. I eventually learned to ignore the noise and focus on the immutable fundamentals: clear workflow design, pristine data quality, and actual user needs. The models will change next week. Those core skills will not.

That you will never feel 100% ready, and that is perfectly fine. Particularly as women, we fall into the trap of thinking we need just one more qualification, one more course, or one more certificate before we are allowed to take up space or offer value. Confidence is built purely through execution. Start. Solve one tiny, real-world problem, share your messy work-in-progress, and the credibility will follow.

I would also add: never underestimate the value of your existing domain knowledge. If you have spent years in marketing, operations, healthcare, or retail, you already know where the real-world friction is. You do not need to completely reinvent yourself. Layer AI on top of the expertise you already have and you create massive leverage. Do not wait for anyone to give you permission to enter this space.

The most lucrative opportunities right now sit perfectly at the intersection of technical capability and deep human empathy. AI Product Operations, workflow consulting, and AI-powered strategy support are massive growth areas. You do not need to be the most technical person in the room from day one. You just need to understand people, untangle messy processes, and define the architecture of a solution.

That operational grit is an incredibly rare and valuable skill set. The industry desperately needs people in AI product management, workflow design, business transformation, and user research to make these models actually usable and safe in the real world. A background in diplomacy, stakeholder management, or operations is not a detour. It is the edge.

AI is absolutely not just for engineers. Yes, the underlying mathematics is complex, but making these tools function safely in a business environment requires communication, judgement, and structural thinking. You do not need to understand how a neural network is trained from scratch before you can build a useful workflow.

Pick one deeply annoying, repetitive bottleneck in your own day-to-day work, use an LLM to solve it, and learn the technical language organically as you go. My own roots are in international economics and diplomacy. I did not grow up as a programmer. Every detour I took, from sports governance to energy infrastructure to Computer Science, turned out to be exactly the preparation I needed.

The most fascinating opportunities happen precisely when you bring completely different experiences together. The most important thing is simply to begin.