SpotWise

Inspiration

Choosing where to open a small business is one of the most expensive decisions a founder makes. The relevant information exists, but it is scattered across census tables, transport datasets, business directories, property reports, and unfamiliar maps.

There was another barrier we wanted to remove: language.

Location analytics is already complex. For someone who does not feel comfortable discussing their business in English, it becomes even harder. They may have to translate their idea, learn technical terminology, and navigate unfamiliar software before they can even begin the analysis.

SpotWise takes a different approach: we do not ask people to translate themselves into the software—the software adapts to them.

We built SpotWise to make evidence-based location advice conversational, understandable, and accessible in the language a founder is most comfortable using.

What it does

SpotWise is a voice-first, multilingual AI location consultant for small-business founders.

Users speak or type to Mel, describing the business they want to open, the customers they hope to reach, and what matters most about the location. Mel turns this natural conversation into structured requirements and analyses locations across Greater Melbourne.

SpotWise provides:

• Ranked suburbs and specific locations to investigate
• An interactive suitability map
• Comparisons between suburbs and individual locations
• Evidence from the 700-metre catchment around a selected point
• Adjustable priorities for customers, competition, transport, activity, growth, and affordability
• Explanations containing figures, data sources, and publication years
• Voice and text consultation in English and Mandarin

SpotWise is designed as a multilingual product. For the hackathon, we focused our implementation and testing on English and Mandarin because of the available time, while building the consultation model so more languages can be added.

SpotWise does not claim to predict business success. It helps founders turn hundreds of possible locations into an evidence-backed shortlist for further investigation.

How we built it

SpotWise combines a conversational AI agent with a deterministic geospatial analysis engine.

Mel is powered by ElevenLabs and uses structured tools to gather requirements, run an analysis, inspect locations, compare alternatives, and control the map. OpenAI supports our backup text experience and generates written consultation briefs.

The AI interprets what the user means and decides which tools to use. Our own scoring engine calculates the recommendations and supporting evidence. This separation means the language model does not invent scores, statistics, or locations.

The analysis combines:

• More than 176,000 business locations
• 16,439 H3 analysis cells
• 747 Greater Melbourne suburbs
• More than 23,000 public transport stops
• Census demographics and employment
• Train patronage and activity centres
• Population and business-growth projections
• Property prices as an affordability proxy

We processed and validated the data using DuckDB, SQL, and TypeScript. We combined City of Melbourne’s CLUE business census with Overture Maps after finding that each source had different coverage strengths.

The application uses Next.js, React, TypeScript, MapLibre, and H3. Its conversation, map, rankings, and manual controls share the same requirements store. A user can change languages or interaction modes without changing the underlying business requirements or losing their consultation.

We also built automated tests covering the scoring engine, agent tools, geographic constraints, state continuity, language handling, and data loading. We verify that the expression colouring the map produces the same scores as the JavaScript ranking engine.

Challenges we ran into

Multilingual conversation involves more than translating an interface. Business descriptions can contain local expressions, mixed languages, proper names, and industry-specific terminology. The system must preserve the user’s actual intent when converting that conversation into structured requirements.

We also needed language preference to remain separate from business logic. Switching from English to Mandarin should change how Mel communicates, but it must never silently change the user’s customers, priorities, selected location, or recommendations.

Because of the hackathon timeframe, we concentrated our live experience and testing on English and Mandarin. Building reliable evaluation with native speakers across many more languages remains an important next step.

The data presented another major challenge. Our sources use different geographic boundaries, formats, classifications, and publication years. Some values are missing, suppressed, or only available through proxies.

Competitor coverage was especially difficult. We evaluated Overture Maps against the City of Melbourne’s CLUE register and found different gaps in each. We therefore use CLUE inside the City of Melbourne and Overture across the wider metropolitan area.

Finally, location decisions contain unavoidable trade-offs. Busy precincts often have more potential customers and more competitors. SpotWise needed to expose those tensions clearly instead of hiding them behind an unexplained recommendation.

Accomplishments that we're proud of

We built a working consultation in which multilingual conversation is connected directly to real geographic analysis.

A founder can explain an idea in English or Mandarin, let Mel gather the relevant context, receive recommendations, explore them on a map, compare alternatives, correct a requirement, and see the analysis update without starting again.

We are particularly proud that:

• Users can discuss their business in the language they are more comfortable using
• Language preference remains independent from business requirements
• The language model does not invent scores or evidence
• Recommendations are reproducible and calculated by our own engine
• Voice, text, map interactions, and manual controls share one continuous state
• Evidence includes its source and publication year
• Data limitations and proxy measures are shown openly

SpotWise demonstrates that advanced location analytics can feel like a conversation rather than a specialist dashboard.

What we learned

We learned that accessibility is not only about making an interface simpler. It is also about allowing people to express complex ideas without first translating those ideas into another language or learning the software’s terminology.

For multilingual AI, technical language support is only the starting point. A production product needs native-speaker evaluation, consistent business terminology, culturally appropriate communication, and reliable handling of mixed-language conversations.

We also learned that trustworthy AI benefits from a clear boundary between interpretation and calculation. Mel is well suited to understanding conversational intent, while a deterministic engine calculates location scores consistently.

Finally, more data does not automatically produce better decisions. Each source has its own gaps and definitions. Validating sources, explaining proxies, and showing uncertainty were as important as designing the scoring formula.

What's next for SpotWise

Our next major step is integrating commercial rent data. SpotWise currently uses residential property prices as an explicitly labelled affordability proxy. Commercial rent, vacancy, lease conditions, and available-premises data would connect suitability with the actual cost of operating in an area.

We also plan to expand beyond Greater Melbourne, beginning with Sydney and Brisbane. This requires local data pipelines, transport sources, planning information, and geographic validation for each city.

Multilingual expansion will remain central to that growth. We want to add and evaluate more languages with native speakers, allowing founders from more communities to describe their businesses naturally.

Future development also includes:

• Commercial-property and available-premises integration
• More languages and multilingual evaluation
• Stronger competitor coverage and entity matching
• Travel-time and pedestrian-activity data
• More personalised explanations for different industries

Our long-term goal is for a founder to open SpotWise anywhere, speak naturally in their preferred language, and receive a clear, evidence-backed starting point for one of the biggest decisions their business will make.

Built With

• next.js
• react
• typescript