Cerebro MVP: County intelligence
Gorden gecko said it best. “the most important commodity I know of is, information”.
-
Introduction
Brief: Build a tool to give companies fast, reliable market insights for international expansion—no more wading through endless PDFs.
Challenge: Deliver up-to-date, relevant info via a chat-based interface that feels natural and transparent.
Goals: Help users get actionable answers, build trust in the data, and make research feel easy.
-
Discovery Phase
Stakeholder interviews: Sat down with founders and early adopters (expansion teams, research leads).
Business requirements: Real-time data, easy updates, and support for multiple company sizes.
User needs assessment: Users want clear, quick answers—preferably through conversation, not document scanning.
-
Problem Framing
Problem statement: Companies expanding abroad can’t find reliable, actionable local intelligence fast enough.
Success criteria: Users get precise, timely answers in a few chat exchanges.
Constraints & assumptions: Needs to feel human, support multiple personas, and keep trust front and center.
-
User Research
User interviews: Expansion leads and SMEs described working “in the blind” and wasting hours on research.
Competitive analysis: Most solutions were static reports or clunky dashboards.
Market research findings: Conversational UIs drive engagement—especially for complex, location-specific info.
-
Synthesis
User personas: “The Outbound Startup,” “The Series B Expansion Team,” “The SME Explorer.”
Journey mapping: From first market question to actionable insight, with trust-building at every step.
Key insights & opportunities: Users want transparency, quick wins, and the ability to ask both casual and business-specific questions.
-
Solution Development
Information architecture: Chatbot core, with support for both casual and business intents.
Wireframes & prototypes: Explored flows for asking questions, getting answers, and following up.
Visual design system: Clean, chat-first UI with clear feedback, simple icons, and context cues.
-
Testing
Usability testing: Ran with target users—flagged trust issues with passive feedback and “robotic” responses.
User feedback: Added active waiting, more context, and gradual info display to boost trust.
Design iterations: Tweaked chat flows, improved error handling, and refined persona support.
-
Implementation
Handoff process: Annotated wireframes and chat flows delivered to devs.
Development collaboration: Weekly check-ins to fine-tune NLP and feedback mechanisms.
Go-to-market strategy: Launched with a pilot group, promoted via international business forums.
-
Results & Impact
Metrics & KPIs: Faster research times, higher trust scores, and more actionable insights per user.
User adoption: Strong uptake among expansion teams and SMEs—users reported “finally getting answers they trust.”
Business impact: Helped companies make smarter, faster moves into new markets.
Lessons learned: Transparency, conversational flow, and active feedback are key to building trust in AI-driven research tools.
Having the ability to gain market insights for a region you are intending to move into gives you a better understanding of how & when to make the correct moves. This is how Project Cerebro was at first initiated. The demand and need for a ‘one-stop’ information portal for companies ambition’s to expand into overseas markets, including those from overseas that have the intention to be present in Singapore.
From the initial interviews conducted we found that.
These companies often find themselves working in the ‘blind’, not having reliable or good information (data) to take the first right steps venturing abroad.
Through Project Cerebro, with the help of AI and data analytics tools, companies can have tools to make better-informed business decisions quickly enabling themselves to make a successful foray into a new market with the latest and most up-to-date information at hand.
When looking at how best to give this information to the user we looked at what currently exists in the market. This primarily consists of large PDF documents that the user must then scan through to get the relevant knowledge that relates to them. This experience can be laborious and time-consuming.
Using the personas as our guide we seen that our users were people that were outgoing and liked to communicate when finding information. This was the definition of our modality in the form of a conversation users could get the information they required. The concept was defined that using AI to scan the documents for information the delivery mechanism would be in the form of a chatbot.
Looking at the rationale for bots was not because of it being the current trend. This direction was made purely on the outcome of what bots are best at doing. Defining intent. Simply put, intents are the intentions of the end-user, these intentions or intents are conveyed by the user of the bot.
In further research, we defined that intent within bot’s is defined in two ways. casual (how are you today) and business (what are the tax issues in Vietnam)
The entire experience is rooted in transparency. People demand transparency from virtually every aspect of life and more so when they are trying to get relevant, locational & context-sensitive information. With the bot’s intent defined, we went about crafting flows on how this would be achieved
After we had a working flow and concept we needed to prove that the hypotenuse was correct that delivering this level of information to users by parsing the data in a more relaxed and easy way. Numerous eye movement tracking sessions were conducted to understand the user’s focal point of concentration.
This helps us see what content do users prioritized, focused on and what gets overlooked. It was clear from the interviews that users had issues trusting the data that was given back to them, as they could feel it was generated from a computer. Elements of the uncanny valley were being displayed from user’s feedback.
This was most appropriate in the user’s reaction to the waiting time. The systems Passive waiting, lost trust in users. The best outcome to gain user’s trust in the information being displayed was to use.
Active waiting interaction gives the user control over the information. This was visualized using a NIP interface.
Natural language processing allows the user at each time that state changes, to have a brief moment of active waiting as they start to process the information presented to them, giving them context about what is eventually going to arrive.
Further user’s feedback from NIP’s gradual display of information, had a better retention rate of effectiveness, efficiency & satisfaction.