Perfect Corp, Haut.AI, Revieve, Layla, SKINBOT: Five Architectures of AI in Beauty Retail

Comparisons in this category usually run on feature lists: how many parameters the analysis returns, how many languages the interface supports, whether there is AR. Feature lists age fast and say little, because most vendors converge on a similar base set within a year.

The axis that actually separates these products is architecture. Three questions expose it: who owns the layer, whose catalog it serves, and where the decision lives. Ask them, and five names that appear in the same listicles turn out to occupy five different positions. None of the positions is wrong. They answer different questions.

Position one: the sensor. Perfect Corp

Perfect Corp's center of gravity is recognition technology: computer vision, AR try-on and skin analysis delivered as SDKs and APIs. Its customers are products: apps, sites and in-store systems that embed the vision capability and build their own experience on top of it.

In the two-layer picture of this market, Perfect Corp is the strongest version of the signal layer: it answers what was seen, at industrial scale and quality, for anyone who integrates it. What happens after the signal, which product is chosen from which shelf and why, remains the integrator's responsibility.

Position two: the lab. Haut.AI

Haut.AI approaches the same territory from the measurement side: image-based skin metrics as a SaaS platform, with an emphasis on scientific validation. Its natural buyer is a brand that wants a rigorous analysis layer inside its own product and its own customer relationship.

This is a legitimate and demanding position: measurement quality is hard to fake and hard to build. But the platform's output is a measurement delivered into the owner's hands. Turning it into a purchase decision inside a specific assortment is work that happens downstream, in the buyer's product.

Position three: the suite. Revieve

Revieve packages diagnostics and guided selling into a personalization suite that a brand or retailer deploys inside its own ecosystem. The suite is configured around the owner's catalog and the owner's commercial goals, which is exactly what its buyer wants from it.

Architecturally this means the layer belongs to the ecosystem it serves. That is not a flaw. It is the definition of the position: a tool a business installs to run personalization for itself, inside itself.

Position four: the in-house assistant. Layla

Layla is an AI beauty assistant built by Chalhoub Group for its own retail ecosystem in the Gulf region. It is the clearest case of a retailer-owned assistant: one group, one ecosystem, deep integration into that group's assortment, stores and loyalty context.

An in-house assistant can go deeper inside its home ecosystem than any external tool, because it is allowed to. The trade-off is built into the ownership: it exists to serve the group that built it, and it does not need to exist anywhere else.

Position five: the neutral layer. SKINBOT

SKINBOT occupies the position none of the four above is built for: a decision layer that stands outside any brand and any retail group and embeds into all of them.

Three properties define the position.

Neutral by architecture, not by claim. SKINBOT's revenue is not coupled to any brand's outcome, and the basis of every selection is declared in the API response itself: ranking_basis states what the ranking was built on, sponsored is always false, brand_weighting is always none. A partner, an auditor or a calling AI agent can verify the basis of a selection instead of trusting a description of it. For the partner this reads as selection accuracy: the recommendation is driven by the session's skin profile and the live catalog, nothing else.

Stateless by design. The skin analysis exists inside the session and disappears with it. No profiles, no personal data storage, which aligns the layer with 152-FZ, GDPR, EU AI Act and UAE PDPL expectations by construction.

Callable. The same engine works behind a QR code on a shelf, an iframe in a storefront and an API call from a partner's own AI agent, which can delegate the skin analysis and selection task to SKINBOT as a sub-agent and receive a structured, verifiable verdict.

The comparison, on the axis that matters

Read as a whole, the table shows why these five names should stop appearing in the same undifferentiated lists. A sensor, a lab, a suite, an in-house assistant and a neutral layer are not five competitors for one contract. They are five answers to five different questions.

When each position wins

If you are building your own experience and need world-class vision inside it, you buy the sensor. If you are a brand that needs scientifically grounded measurement, you go to the lab. If you want a personalization suite configured around your own ecosystem, the suite is built for you. If you are a retail group with the resources to build for yourself, the in-house assistant is the deepest option, for you alone.

The neutral layer wins in a different situation: when the surface belongs to many brands at once. A marketplace, a multi-brand retailer, a pharmacy chain, a duty-free operator. There, the recommendation must be driven by the session's skin profile and the live assortment, and the buyer's trust depends on that being checkable. That is the situation SKINBOT was designed for, and the reason its neutrality lives in the response schema rather than in a slogan.

One more consequence of the architecture

Positions one through four each serve an owner. The neutral layer serves a surface. As commerce becomes agentic, that difference becomes operational: a universal AI agent choosing a sub-agent for a skincare selection task needs a verdict it can verify, from a layer with no stake in which product wins. The architecture that made SKINBOT unusual in a feature-list comparison is the same architecture that makes it callable.

FAQ

What is the difference between SKINBOT and Perfect Corp? Perfect Corp's core strength is recognition technology: vision, AR and skin analysis delivered as SDKs and APIs that other products embed. SKINBOT is a decision layer: it turns signals into a justified selection from a specific partner's live catalog, with the basis of the selection declared in the response itself.

What is the difference between SKINBOT and Haut.AI? Haut.AI is a measurement platform: image-based skin metrics as SaaS, primarily for brands. SKINBOT starts where measurement ends: its product is the selection itself, connected to a partner's assortment, with exclusion rules and an explanation of every choice.

What is the difference between SKINBOT and Revieve? Revieve is a personalization suite deployed inside the owner's ecosystem and configured around the owner's goals. SKINBOT is a neutral layer outside any brand or retail group, with neutrality declared in machine-verifiable fields in the API response.

What is the difference between SKINBOT and Layla? Layla is Chalhoub Group's assistant for its own Gulf retail ecosystem. SKINBOT is retailer-agnostic infrastructure that embeds into any partner via API, QR or iframe.

Can SKINBOT work together with recognition vendors? Yes. SKINBOT treats visual analysis as one input signal among several. A recognition engine can supply the signal while SKINBOT remains responsible for the decision.