How we rebuilt an AI agent SaaS in eight weeks — and helped a brand reach a thirteen million dollar valuation.
LeParfum had everything lined up. A partnership with IFF — the number one perfume creator in the world. A strong marketing and sales team. Customers begging them to launch.
The founder's wife is a perfume expert with years in the fragrance world. Together they shaped a genuinely original idea: luxury, personalized perfume curated by an AI agent. Instead of sitting with a human perfume expert, you talk to the AI — it figures out your taste, sources the right fragrance, and handles everything from recommendation to purchase.
They had spent a full year trying to build the platform. They didn't hire amateurs — these were developers from Google and Facebook. Serious resumes. Serious experience. But being a great software engineer and being an AI expert are two very different skill sets. After a year of iterations, what they had was a pile of code that didn't do what it needed to do.
Most dev teams jump straight into building. We audited instead. We interviewed every team member. We sat with the founders and understood not just what they wanted to build — but what they wanted to be.
Their vision wasn't a standalone tech product. It was becoming a premium perfume brand. Five-star hotels as distribution partners. Ritz Carlton. That kind of level. The AI agent couldn't just "work." It had to speak the right language — tone, vocabulary, the way it guided someone through choosing a fragrance.
This is the core of the whole project — rebuilding the AI agent from scratch and getting it to perform at the level the brand needed.
This was early 2025. AI agents weren't what they are now. There was no playbook. No best practices anyone agreed on. We went all in on agentic AI when most companies were still debating whether to use ChatGPT for customer support emails.
The work was detailed. Getting the tone right — refined and knowledgeable, not robotic. Layering in the brand's marketing tone so the experience was consistent from the ad someone clicked to the conversation with the AI.
Say you're staying at the Ritz Carlton. You buy a personalized perfume through the AI. Along with your order you get a physical metal card with a QR code on it. The card is also NFC-enabled — tap or scan. That card is how you share your perfume with other people. If they buy, you earn a commission. Real money, not points or credits.
A polished interface that matched the brand. An AI agent that sounded right, responded consistently, and delivered the experience their customers expected. A clean, maintainable codebase their team could actually build on.
Enterprise deals closed post-launch: Ritz Carlton Cruises. Wynn Casino (#2 globally, Las Vegas). Foxwoods (#2 in the US). Multiple Premier League football clubs. IFF (#1 perfume creator). Plus others under NDA.
We've shipped agentic AI products since early 2025. If the standard tools don't solve it, we go deeper.