Brolex

• 

  Voice agent displaying product suggested

• 

  Brolex Landing page 1

• 

  Brolex Landing page 2

Inspiration

The spark came from two very different moments. First, while I was with my girlfriend at a bridal boutique, I saw how much personal attention the staff gave every customer, something that feels missing in most online stores. A few days later I heard the CEO of HappyRobot talk about AI voice agents in logistics, and that showed me how powerful real-time voice could be. Those ideas clicked: why not bring the same personal feel of a boutique to an e-commerce site using an AI voice agent? I kept the tone a bit satirical by spoofing Rolex—a brand I admire—as a nod to both craftsmanship and storytelling.

What it does

Brolex is a live voice assistant that sits right inside a watch shop. Shoppers speak to it the same way they would talk to a store associate. The agent greets them, answers questions about models, suggests add-ons like insurance, and can drop items straight into the cart before guiding them through checkout. The back-and-forth feels smooth, so the line between browsing and buying almost disappears.

How we built it

I built the prototype solo, relying almost entirely on Bolt for coding. The agent voice and reasoning come from the ElevenLabs Conversational Agent API, streamed into a Next.js front end. There’s no dedicated backend yet; the agent pulls product data directly from the page and responds in real time through WebSockets. This setup let me stay focused on voice features instead of infrastructure, and it kept latency low enough to feel natural.

Challenges we ran into

Integrate it properly all the tools and re-design a lot the features with a focus in user experince.

Accomplishments that we're proud of

In one weekend I delivered a full voice-shopping flow—browsing, advice, cart, checkout. The ElevenLabs integration stayed stable, and the brand parody landed well without looking unprofessional. Shipping the demo on time felt like proof that voice commerce can be both quick to build and enjoyable to use.

What we learned

Voice agents will enter into e-ccomerce really soon, it is a matter of time for the reduction of the cost of this tools.

What's next for Brolex

Next up is user research to see if eccomerce managers really want voice help. If interest is there, I’ll turn the prototype into a professional app, add lightweight analytics, and test multi-language support.

Built With

• bolt
• elevenlabs
• next.js