Instinct’s new recommendations test whether users want an AI shopkeeper

A human layer on top of AI

Founder Noah Shinn introduced the feature on Tuesday. Instinct says it plans to work with specialists whose “human taste” can add something beyond information gathered from the open internet. The company has not named any partners.

Shinn offered several examples of how the approach might work:

The aim, Shinn said, is to offer recommendations that stand apart from both internet search results and other chatbots’ answers.

That promise depends on more than having experts curate options. The recommendations also need to arrive at the right moment. Several users said Instinct had suggested products they had not asked for, including travel accessories.

Shruti Gandhi, a general partner at AI-focused investment firm Array VC, wrote on X that she did not need a travel suitcase, sunglasses or a wide-brimmed hat. She would rather have received a personalized shopping list of things she actually needed.

Entrepreneur Andrew Yeung also said on X that he had received unsolicited recommendations. Chat Jogalekar, founder of a platform for buying businesses, said the suggestions were the first time he had felt repelled by an AI assistant. Instinct had drawn on his emails and upcoming trips to recommend items including a travel mug, apparently because he subscribes to Blue Bottle Coffee, as well as other travel accessories.

Relevance is not enough

Recommendations may feel helpful when a user asks for them or when they fit naturally into a conversation. Sent without that context, even a well-matched suggestion can look like an ad. Some users said Instinct should ask permission first and let people choose which sources it uses.

My read is that the feature’s central problem is consent, not curation. A list assembled by a local expert may be better than a generic search result, but that does not answer whether the user wanted a list at all. The announcement is quiet about how Instinct will decide when to surface suggestions, or whether users can turn them off.

That question matters as Instinct competes with AI agents from Meta and OpenAI, which users say do not send similar recommendations. If Instinct delivers suggestions at random, they will need to be sharply relevant to avoid feeling like spam.

Commerce behind the recommendations

Instinct has not said whether it already earns money from recommendations or plans to do so. The feature could support advertising, commerce or affiliate links, but the company has not disclosed a business model for it.

The launch follows Instinct’s confirmation of a $1 billion Series C round that valued the company at $10 billion. Instinct has not disclosed how many people use its product, leaving one important measure of that valuation unknown.

In a recent podcast, Shinn said the agent regularly makes transactions for users. In an interview with investor Patrick O’Shaughnessy, he said annual transaction volume through the platform was approaching $1 billion, with more than 50% of transactions related to travel.

I think the key tension is that transaction activity may make recommendations more valuable to Instinct precisely when unsolicited pitches risk making the assistant less welcome. The company’s next challenge is to show that its human-curated layer can earn a place in users’ routines before commerce becomes the part they notice most.