Context
TopZaklad connects people looking for a place to go with venues that want to be found. Users discover restaurants, cafes, and bars that match their mood and occasion. Business owners get tools to manage their presence and attract clients.
The product was built in 2 phases: first, a web catalog solving a visibility problem in small Ukrainian cities. Then a full pivot to an AI-powered iOS app targeting major cities. I designed both from scratch.
Business model: monthly B2B SaaS subscription for venue owners, covering profile management, analytics, and extended features. Promoted placement in search and catalog as additional paid options. Free for users.
Challenge 1
Traveling across small Ukrainian cities, people can't find venues - Google Maps is empty or outdated there. Businesses exist but are invisible to potential guests. Owners have no tools to reach their local audience.
Goal: Build a web platform where users can find venues, owners manage their presence and get tools to grow.
Solutions
We conducted a survey before building anything and got some key insights:
- Google Maps data for small Ukrainian cities is either empty or years out of date. The only way to fix that was to build the database manually. So we did!
- Users don't need a complex discovery experience. They want to find what's available nearby, fast.
- For business owners, the most important thing is to have an affordable promotional opportunity.
We built a catalog with map integration, and basic tools for owners to manage their profile and run promotions.
Phase 2
- The web platform was built but six months later, the capabilities of AI have grown significantly. It was decided to shift the product's focus and develop a mobile app based on OpenAI, what previously required a large team and a substantial budget became possible in much less time and at a much lower cost.
- The product was reimagined: focus shifted from small cities to major ones, from web to a mobile app powered by AI search.
Challenge
In big cities, choosing a place to meet is a real headache. 50% of people keep going back to the same places because they're afraid of being disappointed. The search takes 15-30 minutes, and even then, the result often falls short of expectations.
People often don't even know what they want. Filters assume you know exactly what you're looking for: "restaurant, Italian cuisine, under $20." AI works even when you don't know: "I'm 25, I want to have a great night out with friends, surprise me!"
From a business perspective, the mobile app has unlocked significantly greater potential for monetization and scaling.
Solutions
1 - Onboarding
AI-powered venue search is a new pattern people need to see it before they commit to signing up.
The onboarding shows the product first: four screens covering AI Search, Suggestions, Catalog, and Lists. Optional Sign Up comes at the end. The idea is that a user who's already engaged with the product has a real reason to create an account.
We intentionally skipped a personalization step in onboarding. At MVP stage, reducing friction matters more than collecting preference data with nothing to validate it against yet.
2 - AI Search
Problem: Filters have a fundamental limitation. It only works when you already know what you want. If you're looking for "Italian, under $20, near me" - filters are fine. But if you just want a good night out and have no idea where to start, filters give you nothing useful.
Hypothesis: The bet was that natural language input solves this. People already talk to ChatGPT every day, so the pattern feels familiar. And there's something filters simply can't do: read user reviews and comments to understand context. AI can.
Solution: We built a chat based search with example prompts to help users get started.
AI pulls from the venue database - descriptions, tags, and user comments, and returns a ranked list matched to what the person actually asked for.
3 - Suggestions
Not every user is ready to type a message to an AI. Some people prefer filters. Some just want to browse.
So the app has 3 ways in. AI Search for people who want fast, flexible, conversational input. Suggestions, which is also AI-powered but works passively, it watches what the user likes, saves, and searches for, and builds a personalized venue list.
4 - Lists
During user interviews, people kept mentioning the same thing: they save places in notes, messengers, and Google Maps, but it all gets mixed together.
So we built Lists with custom names and emoji. For each occasion user can create a new collection.
What was delivered
- Web platform: public catalog + B2B owner cabinet (profile, advertising, analytics).
- Mobile app: full design across all flows, Ready for dev.
- Both products designed in ~100 hours alongside full-time work
What I'd validate first
- The most interesting question is how people actually use AI search in practice. What do they type in their first prompt? Do they understand what to write without any guidance? How many tries does it take to get a result they're happy with?
- The second thing is AI search vs catalog usage. Which search way do more people choose, and which one leads to better outcomes like saving a venue or coming back the next day.
- Would AI matched venues get better reviews than filter picked ones? That would show how well the AI actually performs, and give real data to train and improve it over time.