The Interface Dissolves. The Decision Layer Doesn't

Beauty AI is disappearing into messengers. The real question is who owns the recommendation logic underneath

This morning a launch announcement scrolled past me, and I read it twice, because it was quietly proving an argument I've been making for a year.

ERHA Clinic, a dermatology brand from Indonesia with 26 years of clinical history, launched AI skin analysis inside WhatsApp. Built with Revieve, covered by Cosmetics Business. No app to download. Three minutes from first message to personalized recommendation. Skin analysis delivered in the channel where Indonesian consumers already spend their day.

The industry read this as a distribution story. I read it as a confirmation of something structural.

The interface is losing

For a decade, beauty tech competed on interfaces. Better quiz. Smoother AR mirror. A prettier app. The assumption underneath: the brand that owns the most engaging surface owns the customer.

That assumption is dying in public. Beauty AI is not moving into WhatsApp because WhatsApp is interesting. It is moving there because the interface stopped being the moat. Consumers didn't want another surface. They wanted answers inside the surfaces they already had. So the interface layer is doing what interfaces do under pressure: it's dissolving. Into WhatsApp. Into Telegram. Into the chat window of whatever model a person already talks to at midnight.

This is the shift I call the callable layer: intelligence stops being a destination you visit and becomes a function any channel can call. The ERHA launch is a textbook example. The skin analysis isn't an app anymore. It's a capability living inside a messenger.

If you build in this industry, the direction of travel is now explicit. The question is no longer whether your tool will live inside someone else's channel. It's what your tool will be when it gets there.

What dissolves and what doesn't

Here is the part the announcement doesn't say, and the part that matters more.

When the interface dissolves, two layers separate that used to look like one product.

The first is the conversation layer: where the interaction happens. WhatsApp today, a model's chat window tomorrow. This layer belongs to platforms, and brands rent it.

The second is the decision layer: the logic that turns a person's skin, context, and constraints into a recommendation. This layer doesn't dissolve. It just becomes invisible, which makes its ownership more consequential, not less.

And look at who owns it in this launch. The recommendation engine is built around one brand's catalog, one brand's protocols, one brand's 26 years of expertise. Inside ERHA's world, that's exactly right. Clinical credibility plus first-party data plus a familiar channel is a strong formula, and I mean that without irony.

But run the Indifference Test on it: would this system ever recommend a competitor's product if it were genuinely better for this specific person? It may be useful, clinically informed, and well executed. But it is not indifferent. It is not designed to arbitrate between catalogs. Structurally, it is there to translate one brand's expertise into a more accessible advisory experience, and it does that well.

The invisible shelf, now in your messenger

This is where the category quietly changes shape.

When recommendations lived on a retailer's website, you could at least see the shelf. Sponsored placements were labeled. The store's logic was visible in the layout. Now the shelf is a chat bubble. There's no aisle to walk down, no adjacent products in your peripheral vision. Whatever the system says is, functionally, the entire shelf.

I call this the invisible shelf, and every brand-aligned tool that moves into messaging makes it a little more invisible. The consumer experience improves. The consumer's field of view narrows. Both things are true at once, and the industry only talks about the first one.

Multiply this launch by every major beauty group, because they will all do this, and you get a landscape where each brand operates its own closed advisor inside the same messengers. A consumer choosing between brands gets no help from any of them, because between-brand choice is precisely what none of these systems are built to serve.

The unclaimed layer

So the real map after this launch looks like this:

The conversation layer is claimed. Platforms own it.

The brand advisory layer is being claimed right now, launch by launch, and the ERHA case shows how well it can be done.

The layer between brands, the one a consumer would trust when the question is not "which ERHA product" but "what should I actually use", is still unclaimed. It can't be built by any single brand, for the same reason a courtroom can't be run by one of the litigants. It has to be neutral by architecture, not by promise: no retention of a favored catalog, no revenue coupled to a specific recommendation, no write-access for brands to the rubric that judges them.

That is the layer I have been building toward: not another beauty interface, and not another brand assistant, but a neutral decision layer that can be called by any channel and still remain accountable to the person making the choice. Launches like this one don't compete with it. They create the conditions that make it necessary. Every brand-aligned advisor that gets better makes the absence of a neutral one more visible.

The interface war is ending. The interfaces lost, gracefully, by dissolving into the channels people already love. What's left standing is the question the industry hasn't answered yet: when every brand has an advisor in your messenger, who speaks for the choice itself?

What I'm thinking about next: what happens to a brand's pricing power when the invisible shelf is run by a system it can't pay. If you have a view on this, reply on Substack, I read everything.