Hidden gold

Based in Singapore, this client sells over-the-counter health products across Southeast Asia. While its direct customers are retailers and wholesalers, the real demand is driven by the end-consumer in supermarkets, pharmacies, and online. Putting the consumer first is vital to their success.

The company regularly commissions in-depth consumer research to understand local needs and preferences. Typically, this means paying third-party agencies to conduct one-to-one interviews in local languages across the region. The transcripts of these interviews are turned into consumer profiles that shape brand strategy, product development and marketing.

With multiple studies commissioned by various business stakeholders every quarter across different categories and markets, the company was sitting on a mountain of information with no easy way to extract relevant insights. Even with well-tagged documents, it would be extremely time consuming for anyone to sift through this data for that golden insight they needed.

This problem was self-perpetuating. If someone needed an answer to an important question, the only sensible approach was to commission yet another study, knowing the information could already be hidden in an existing report, but powerless to spend the additional time digging for it.

A custom-built solution

Our customer desperately needed a way to reduce the cost and time involved in conducting new studies when its existing data had the potential to answer pressing queries. We designed and built a bespoke AI-powered solution that our client can use to retrieve relevant insights in seconds.

The tool features a natural language interface that makes it easy for users to interrogate the company’s library of existing consumer studies, generating an answer with source references to validate the accuracy of the response. Users can also filter searches and responses by brand, country and age group. They can even ask and receive answers in multiple Southeast Asian languages.

Why a generic solution isn’t enough

Could you just load all this into a generic AI assistant tool? You may be asking this question as you read this blog, but the reality is that an off-the-shelf solution wouldn’t produce responses anywhere near as good.

A solution like Co-pilot can’t be finely tuned to search multi-language transcripts, apply tailored filters, and return answers with clear citations that stand up to scrutiny. Crucially, there’s also no mechanism for feedback and iteration – something that helped us immensely during testing to hone the responses to be as relevant and useful as possible.

By designing a custom solution, we ensured the system could handle complex, nuanced data while delivering business-ready responses in seconds – something a generic plug-and-play tool simply can’t match.

Faster, richer, and more cost-effective insights

The benefits have been immediate. The client now spends less on repetitive research and gets answers in moments instead of weeks. Each new study doesn’t just provide a point-in-time snapshot, it adds to a living knowledge base that anyone can access.

An important part of this success is the ability to fine-tune how the system works. By adjusting parameters like the weighting of keyword versus vector matches, how much context is retrieved, or even how conversation history is handled, the company can ensure the model’s answers are not only fast but genuinely useful and accurate. This level of customization means teams trust the output and feel confident applying insights to real business decisions.

Most importantly, this approach keeps the consumer front and center in every decision without wasting time, money, or valuable research.

A broader lesson for all businesses

Every organization is sitting on a wealth of data and potential insights. But it’s only valuable if it’s accessible and usable. The strength of a custom tool is that we can pair AI’s ability to extract insights from unstructured data with the flexibility to shape those insights for the customer’s specific use case.