AI-Powered 3D PWA with RAG Capabilities for an SGX-Listed F&B Group (SGX: F99)

The brief

Beverage brands compete for attention in a few seconds on a shelf and a few seconds on a phone. A product page with a static photo and a block of copy rarely holds that attention. And more of the questions consumers ask about a drink now go to search engines and AI assistants rather than to the brand’s own website.

The brief was to give the brand a digital experience that people want to spend time in, that answers their questions intelligently, and that keeps the brand visible and accurately described wherever those questions are asked.

We treated those as three parts of one product rather than three separate deliverables. The 3D experience draws people in. The AI layer answers what they want to know. Answer engine optimisation makes sure the same facts reach people who never open the app.

1. An interactive 3D experience

The centrepiece of the app is real-time 3D product rendering. Instead of looking at a fixed photograph, visitors can turn the product, look at it from any angle and explore interactive visual scenes built around the brand.

Everything renders in the browser with WebGL, so there is nothing to download from an app store. The challenge with 3D on the web is rarely making it look good on a powerful laptop. It is keeping it smooth on a mid-range phone over a mobile connection. We built the experience around that constraint, so the scenes load progressively and the essential product information never waits on the 3D to finish.

The 3D work serves engagement, not decoration. Each interactive moment was designed to show something about the product that a flat image cannot, and to lead naturally to the next thing a visitor might want to know.

Read more about our 3D and interactive web experiences.

2. Generative AI with retrieval-augmented generation

The app includes an AI assistant that consumers can ask about the brand and its products. A general-purpose chatbot is a risky thing to put in front of a listed company’s consumers. It can confidently invent ingredients, misstate nutritional information or wander into topics the brand should never comment on.

Retrieval-augmented generation (RAG) addresses that. Before the model answers, the system retrieves the relevant passages from a curated knowledge base of brand and product information. The model then answers from that retrieved context, rather than from whatever it happens to remember from its training data.

In practice, this means a consumer asking about a flavour, an ingredient or where the brand comes from gets an answer grounded in information the brand stands behind. When the knowledge base does not cover a question, the assistant is designed to say so rather than guess.

Keeping the knowledge base separate from the model also keeps it maintainable. When a product launches or information changes, the brand updates the source content and the assistant’s answers follow, without retraining anything.

For a plain-English explanation of embeddings and retrieval, read how vector databases work, or see how we build custom AI assistants.

3. Answer engine optimisation (AEO)

Many people will never open the app. They will type a question into Google, or ask ChatGPT, Gemini or Perplexity, and read whatever answer comes back. Answer engine optimisation is the work of making sure the brand’s own accurate information is what those systems find, understand and cite.

The AEO work on this project covered three layers.

1. Structured data: Schema.org markup that describes the organisation, products and frequently asked questions in a form machines can read without guessing, kept consistent with what is visible on the page.
2. Semantic content architecture: Content organised around the questions people actually ask, with each answer stated clearly and early, and each page focused on one subject so it can be quoted accurately.
3. AI search optimisation: Making the content crawlable and accessible to AI platforms as well as traditional search engines, so the brand is discoverable wherever the question is asked.

AEO and RAG reinforce each other here. The same curated brand facts that ground the in-app assistant also feed the structured, public content that external AI systems draw on. Consumers get consistent answers whether they ask the brand’s assistant or someone else’s.

Learn more about our AI SEO and AEO strategy work.

4. A Progressive Web Application

We built the experience as a Progressive Web Application. A PWA is a website that behaves like an installed app: it can be added to a phone’s home screen, opens full screen and loads quickly on return visits because key assets are cached on the device.

For a consumer brand, that combination matters. A campaign link or QR code opens the experience instantly in the browser, with no app store step to lose people at. Visitors who want to come back can install it in one tap.

The architecture is responsive and mobile-first, because most visitors will arrive on a phone. The same codebase delivers the experience on phones, tablets and desktops, which keeps maintenance in one place rather than spread across separate iOS, Android and web builds.

See our work on websites, web apps and cloud.

5. A connected digital brand strategy

The fifth deliverable ties the other four together. Generative AI, immersive technology and search optimisation are each useful on their own. Combined around one brand, they cover the whole journey from first discovery to repeat visit.

• Someone asks an AI platform about the brand and gets an accurate answer drawn from content the brand controls.
• They follow a link or scan a code and land in an immersive 3D experience that loads in seconds.
• They ask the in-app assistant a question and get an answer grounded in approved brand information.
• They install the app and come back, without ever visiting an app store.

Each step strengthens the brand’s digital presence, and each is built on the same source of truth about the brand and its products.

What this means for listed consumer brands

Listed companies have more to lose than most when a digital channel gets the facts wrong. An AI feature that invents product claims is a reputational risk. A beautiful 3D site that stalls on a phone wastes campaign spend. And a brand that is invisible to AI search is ceding the answer to whoever else is talking about it.

This project shows that those risks can be designed out rather than accepted. Grounding AI in curated brand content, building 3D for real devices and structuring content for answer engines make immersive, AI-powered experiences appropriate for organisations held to a high standard of accuracy.

For a related look at brand identity, AEO and reputation monitoring in the beverage sector, read our AEO case study for a regional beverage brand.

If you are planning an immersive web experience, an AI assistant your brand can stand behind, or better visibility in AI search, speak with Web Wizards SG.