Instinct Tests Push Suggestions Based on Email and Trips
Instinct has launched Instinct Selections, a feature that pushes product suggestions to users without a request. Founder Noah Shinn announced the debut on Tuesday. TechCrunch
Selections is intended to curate personalized lists for dining, travel and shopping. The company says the lists draw on local chefs, designers, architects and travel guides, meaning human experts rather than only software. It did not share who those curator partners are.
Early user reports describe the feature as proactive rather than request-driven. Array VC general partner Shruti Gandhi said she received unsolicited suggestions for carry-on luggage, sunglasses and a brim hat. Entrepreneur Andrew Yeung confirmed he received product recommendations he had not asked for. Chat Joglekar said Instinct suggested products to buy based on what was in his email and upcoming trips.
The launch came shortly after Instinct confirmed a $1 billion Series C financing. Instinct did not share whether it is currently making money from the recommendations or plans to. Shinn has said more than 50% of transactions on the platform are travel-related. Skift
The permission question is where this gets tricky for builders. Pull is simple. The user asks, the system retrieves. Push is harder. The system infers intent from persistent context, such as inbox contents and trip plans, and interrupts with a suggestion. That changes what permission is needed, how many wrong guesses users will accept, and what a miss costs.
The broader context here is the shift toward proactive agents that monitor signals and surface next actions without an explicit prompt. For travel and commerce, the appeal is clear. Packing lists, gear replacement, restaurant shortlists and itinerary add-ons can be assembled before the user thinks to ask. The failure mode is also clear. A suggestion the user did not request reads less like assistance and more like advertising, especially when the data source is email.
In my view, the human-curator element is an attempt to close that trust gap. Algorithmic advice that arrives uninvited carries low credibility. A named chef or travel guide carries more. Without named partners, though, users see the personalization and the push but not the provenance behind the pick.
Worth flagging, the inbox detail deserves close attention from builders. Email ingestion, meaning the assistant reads and stores email, gives high-signal context on bookings, receipts and plans. It also mixes private context with commercial action. Users may accept an assistant reading email to answer a question. Having it read email to initiate a sale is a different contract. That contract needs explicit controls for data scope, retention and opt-out, presented at the moment of use rather than buried in settings.
Looking at what this means for product teams, the monetization question will shape how Selections is interpreted. A curated list can be a retention feature, a conversion funnel, or both. Until the business logic is stated, users will supply their own explanation. In my experience watching my kids adopt new assistants, that default explanation skews skeptical once money might be involved. Stating plainly whether placements are paid, whether clicks convert to revenue share, and whether purchase history feeds future targeting would remove ambiguity.
The optimistic case here remains intact. Travel-heavy assistants have a natural path into contextual commerce, because trips create dense, time-bound needs. A system that knows the itinerary, the climate and the baggage constraints can reduce search labor in a useful way. Done with consent, transparency and reliable curation, push commerce could save time rather than extract attention. The early reaction suggests Instinct has the underlying signal. It now needs the interaction contract to match.
