AI Across Three Roles Without Losing the Human Touch

I'm Dragana, and I am a co-founder of Uradi Zaradi, a two-sided marketplace connecting customers with people who provide household services like cleaning and ironing. Think of it as an Airbnb for small household tasks. We launched in 2016 while we were still students, which meant we had no money, no experienced advisors, and no map for building this kind of platform in Serbia.

Everything we built in those early years came from our own skills, bootstrapping from other jobs, and learning by doing things ourselves. Alongside Uradi Zaradi, I also started another startup called Rinzi, and I work as a digital development advisor and mentor to entrepreneurs, women founders, and social enterprises. But Uradi Zaradi is where it all began.

The challenges were layered. First, we were students with no business-building experience and no one around us who had built a marketplace before. The mentors we found knew traditional services but not digital platforms or online payments.

Second, the market itself was not ready. People in Serbia in 2016 mostly did not pay online, did not trust strangers for household work, and had no framework for understanding what we were offering.

We had to educate customers, build trust from zero, and navigate legal and tax structures nobody had mapped out for a business like ours. To understand the experience from the inside, my co-founders and I went and did cleaning and ironing jobs ourselves, so we could learn directly from both sides of the platform.

We did not build the full product to test the idea. Instead, we created two landing pages, one for customers and one for taskers, to simulate the entire process and see how people responded. We ran Facebook ads and waited.

The first customer paid two weeks after launch, which was more than we expected, given how much fear we had about whether people would trust the service at all.

That moment created a belief in me that this market genuinely needed what we were building, and that belief has never changed. We kept learning from every customer and every tasker, matching people through a Facebook group in the early days before the full technology was in place.

We bootstrapped completely. Family and friends did not understand what we were doing and were not in a position to invest. Banks were not an option at the early stage. So we worked other jobs, used the income from those to cover what the platform needed, and built everything ourselves.

Having a developer as a co-founder meant we did not need to pay for development, and having someone focused on marketing in-house meant we kept those costs internal too.

The only external investment we ever received came in 2019 from Inspire Group, a small amount that helped us integrate a proper payment system and scale for a few months. That is the full investment story of Uradi Zaradi. Everything else came from our own work.

I would do the whole thing the same way, but faster. Every decision we made, every business model we tested, every mistake we went through was part of understanding the market and building something real. I do not regret the process.

What I would change is how quickly I found the right advisors, people who could have helped me move through certain stages without spending months figuring things out myself.

For anyone starting now, my advice is simple: just start. Do a one-pager. Pitch someone. Go to events. Do one thing today. Because you will never know if it is possible until you try, and you can only say you genuinely tried once you have actually done something.

I use AI at several levels. For my own productivity, I use Claude and other language models to handle tasks that are repeatable and time-consuming, like drafting emails, structuring documents, and supporting strategic thinking.

In my role as a digital advisor at a foundation, I also help colleagues and organisations understand how to adopt these tools responsibly, because the adoption rate in most organisations is still very low, often below 1.5%.

The main barrier is not willingness, it is understanding. People are unsure about data privacy, unsure what to delegate, and unsure where to start. Part of my work is helping them see that AI is not there to replace their judgment but to free them from the tasks that do not require it.

It would have saved me significant time, especially in the areas where we struggled most: legal structures, email templates, process documentation, and finding information. All of that took enormous energy in 2016 that AI could now handle in minutes.

But I also believe that going through that process manually gave me something that I could not have gotten any other way: a deep, hands-on understanding of how those processes actually work.

That knowledge is now what allows me to help other entrepreneurs optimise their own operations with AI. If I had skipped the hard version, I would not have the expertise to guide others through it. So the answer is: yes, it would have helped, and also the struggle was necessary.

AI can help with a lot of what surrounds mentoring: research, structuring advice, preparing materials, and following up. I use it for those things. But what I will never delegate to AI is the human part itself: the listening, the challenge, the relationship, and especially that moment when a founder realises something important about themselves or their business.

That moment is irreversible. It changes how someone sees their path. No model can replicate it because it comes from real presence, real trust, and years of shared experience. AI is a tool for efficiency. Mentoring is something else entirely.

I love technology, and I believe it should impact people's lives for good. But right now, the conversation about ethics in AI is too narrow. It is mostly happening among developers and technical people, when it should involve a much wider group, including people with backgrounds in law, psychology, education, and social impact.

The European AI Act is coming into effect, and the GDPR implications are growing, but regulation alone is not enough if the ethical thinking is not embedded in the design of the models themselves. I am actively interested in contributing to that conversation, not just as a user but as someone who works with organisations on digital adoption and understands where the gaps between technology and human reality actually sit.

Start by understanding what tasks in your day are repetitive and time-consuming but do not require your unique judgment. Those are the first things to hand to AI. Then experiment: try different tools, understand their capabilities, and see what actually saves you time versus what creates new complexity.

Do not try to adopt everything at once. Choose one or two tools, learn them properly, and build from there.

The broader principle I believe in is that data and information are the foundation of good decisions at every level. AI gives you access to more of both, faster. But you still have to ask the right questions and know what you are looking for.

That judgment is yours, and it comes from experience. So keep building yours, even as the tools get better.