An AI skincare concierge trained on 30 years of formulation expertise
A premium DTC skincare brand, founder-led, 30-year formulation heritage. · Phases 1 + 2 shipped March 2026.
A premium direct-to-consumer skincare brand needed to scale expert consultations without scaling headcount. We built a multi-channel conversational agent that feels like talking to the founder.
The situation
The brand's customers — women 25 to 45, shopping premium price points — wanted guidance, not a product catalog. Every purchase was a consultation. The founder had built a 30-year knowledge base of formulation science, skin-type interactions, and ingredient pairings that lived in her head and in four binders.
Customer service couldn't scale. Wait times stretched. The team kept turning down TV appearances because there was no one home to field the traffic spikes.
What we built
A conversational concierge that speaks across web chat, SMS, and WhatsApp — wired into Shopify for order history and backed by a vector store of the founder's expertise. The agent:
- Pulls in skin-consultation history per customer, so follow-up conversations remember context.
- Explains why a product fits — ingredient chemistry, not marketing copy.
- Escalates to a human when confidence is low. No fake expertise.
- Runs local inference where possible. Zero per-conversation subscription fees.
The shape of the engagement
| Phase 1 | Architecture, knowledge-base ingestion, Shopify integration |
| Phase 2 | Web chat + SMS rollout with HITL review queue |
| Phase 3 | WhatsApp, multilingual support, proactive re-engagement |
Outcome
- About $25–40k a month in revenue lift from recovered CS throughput and higher-converting consultative flows.
- 24/7 coverage without expanding the support team.
- Zero subscription fees per conversation — the economics improve as volume grows.
What surprised us
The agent was too good at recommending. The first batch of conversations felt like sales calls because every exchange ended with a product. The founder caught it in the review queue and was firm: real consultations sometimes end with "you don't need another product right now, give the one you have eight weeks." We tuned the agent to recommend nothing in roughly a third of conversations. That's when the trust scores moved. Customers started telling the agent things they would never tell a salesperson — bad reactions, breakouts before weddings, twenty years of trying products that didn't work. The agent's job stopped being "sell well." It became "listen first, then earn the recommendation."
The honest version: AI won't replace the founder's taste. It will ensure every customer gets a thoughtful answer at 10pm on a Tuesday — and a warm hand-off to the human when it matters.