Strategy Briefing (Crabstone)

The Algorithm Cannot Recommend a Room

AI has commoditized product discovery, so The Outset and Nécessaire are spending on book clubs, floral workshops and art-world breakfasts. The one thing an algorithm cannot recommend is a room full of people.

Sir John Crabstone

The most defensible asset a beauty brand can now own is a room full of people. This in a year when three in four consumers use large language models to research beauty, wellness or longevity, and a quarter name AI their primary source for shopping research, in BCG and WWD’s survey of 5,000 US beauty shoppers. The machine has taken discovery. The interesting brands are letting it. Visibility, once free, is now a subscription with no ownership at the end of it.

Their rivals are doing the opposite, and doing it well. The industry has coined a term for it: generative engine optimization, the craft of being named by ChatGPT before a competitor is. Brands now budget for it the way they once budgeted for search-engine optimization, and for the same reason: ignoring it costs more than doing it. It is the correct instinct, which is exactly the trouble. A model does not spread attention; it narrows it, and it has already chosen its names. eMarketer’s visibility index puts La Roche-Posay in 81% of facial-skincare queries. You cannot out-optimize a result that is already 81% decided.

What every brand can optimize for, no brand can own.

So The Outset, Scarlett Johansson’s skincare line, opened a book club. Its first title is the novel June Baby, sold beside the skincare in a $100 bundle. In March it took over Fuze House, a studio known for heated sculpting and pilates, for branded class days in New York and Miami; on May 28 it charged $135 a head for an evening of flowers and wine in Greenpoint. Note the direction of the money. The model gives discovery away — The Outset sells the room back at $135, with a florist attached.

Nécessaire went to the art world instead. It sponsored the VIP breakfast at Frieze New York and commissioned an artist to build floor cushions shaped like giant hands. The hand wash it had launched the day before was not the point. The brand called the artists, the creators and the gallerists in the room “a new tastemaker for us.” Taste is the recommendation no model returns; it is granted by people, and people do not scale.

The logic is sounder than it looks. An algorithm can match a complexion to a serum; it cannot supply a reason to care which brand sold it. Community supplies that reason: it places a purchase among particular people, in a room, on a date, with names attached. A customer who found her book club through The Outset has a reason to stay that no coupon can manufacture, and no chatbot can talk her out of.

None of this scales, which is the point. A room like this stays small on purpose, and no rival can buy that intimacy in bulk. The same instinct that makes a brand chase visibility in a language model is the instinct it now has to resist: bigger reach, thinner loyalty. A model can recommend the serum and the bundle. It cannot recommend the company.

Related Coverage