Instinct's unsolicited shopping tips put Noah Shinn's commission model to the test
Instinct users described recommendations tied to earlier conversations; the messages were not labeled as ads, and no evidence shows they generated commissions.
By Ryan Merket · Published
Primary source: Business Insider
Why it matters
Instinct's commercial model depends on acting from personal context. The reports raise questions about when it should make proactive recommendations, how users can control them, and whether merchants pay when users buy.
Three Instinct users told Business Insider that the personal AI agent sent them shopping recommendations they had not requested. Shruti Gandhi and Tanmaye Bhatia described suggestions tied to details from earlier conversations; Girum Tihtina said he could not connect his recommendation to a prior discussion. The reports, published by Business Insider on October 5th, come as founder Noah Shinn (@noahrshinn) builds a business around helping users act on personal context and potentially earning commissions when they buy.
Instinct says its assistant connects to a user's applications and devices and can act through text, calls and email, without requiring another app. Shinn's earlier work focused on agents that use feedback and memory to improve their decisions: as a Northeastern undergraduate, he coauthored "Reflexion: Language Agents with Verbal Reinforcement Learning", then joined Sierra as an early employee, according to Fortune. He is now asking users to trust an agent with the context that can make everyday assistance useful. Shopping suggestions based on that same context can feel different when the user did not ask to shop.
When help turns into a pitch
Shruti Gandhi, a general partner and founding engineer at Array Ventures, said she had not asked for shopping help, Business Insider reported. A screenshot she shared on X shows Instinct drawing on an upcoming conference and keynote to produce a five-item list, including a $250.75 carry-on suitcase, sunscreen and a power bank.
Tanmaye Bhatia, AI lead at Primary Venture Partners, said he received suggestions including $15 golf alignment sticks after earlier discussions about his golf swing and travel. In a screenshot he posted, the agent framed the picks as being "for the range and the road." He told Business Insider he did not think the message was an ad, but called it "a bit of a turnoff" and said he would have preferred Instinct to ask before making occasional recommendations from his personal history.
Girum Tihtina, cofounder and CEO of medical-tourism marketplace Doctours, said a message recommending a "first office kit" arrived when he had not been discussing shopping. It included a monitor he valued at roughly $200. "I have no idea why it chose what to show me," he told Business Insider. He said he wondered whether Instinct was experimenting with ways to make money, while acknowledging that he could not identify why the agent selected those products.
The messages were not labeled as ads. Business Insider could not establish whether Instinct earned money from any resulting purchases. The reports therefore do not show that the recommendations were paid placements or that Instinct received a commission.
The business model meets the product promise
On September 28th, Instinct announced a Shopify partnership, expanding its shopping tools. The next day, Shinn began rolling out Instinct Selections, a feature he described as drawing on recommendations from chefs, designers, architects and local guides.
In an Invest Like The Best interview published September 28th, Shinn described a model in which Instinct remains free for users while merchants pay it a cut for bringing them customers and helping complete purchases. He also said he did not want Instinct to influence a user's behavior toward buying something the person did not want. His comments do not establish whether the specific recommendations came from a commercial arrangement, an experiment or the agent's own attempt to be useful.
Marketing experts quoted by Business Insider said the screenshots do not prove covert advertising. If Instinct earns money when its suggestions lead to purchases, a commission can reward completed transactions. A proactive agent can also create shopping occasions by deciding that a user's travel, hobbies or work plans call for buying something.
That differs from a user searching a store or asking a chatbot for product options. Instinct's pitch rests on remembering what matters to a person and acting without requiring a fresh instruction for every task. For recommendations, that raises a question of control: when should the agent take the initiative, and when should it ask first? Bhatia suggested getting consent before turning personal history into recurring product advice.
The Reflexion paper explores how language agents can use feedback and memory to guide future actions. Instinct's commercial proposition also depends on accumulated context. The more relevant that context becomes, the more valuable the assistant may be, and the more important it is for users to understand when that context is being used to serve them and when it could benefit Instinct or a merchant.
Instinct's September 28th $1 billion Series C at a $10 billion valuation, backed by Sequoia, Benchmark and Coatue, gives Shinn capital to build a personal agent as Instinct expands access and product features. RuntimeWire reported the round last week. As Instinct grows, its economic model will need to preserve the trust the product requires.
RuntimeWire has also covered user reports of capacity warnings and a separate disputed screenshot involving another person's purported financial details. Those reports concerned different issues and do not establish a pattern of misconduct. They raise separate questions about capacity, personal data and unsolicited recommendations as the agent moves from handling tasks to acting proactively across a user's digital life.
Instinct's website describes an assistant that can follow up on dropped threads, arrange airport rides and book a handyman. Shopping fits the same promise of reducing effort. Users need a way to tell when a recommendation is based on their preferences, whether a merchant pays for the transaction, and how to stop unsolicited suggestions. None of the three accounts establishes that Instinct crossed into paid advertising. They do show that, for a product built on personal context, even a useful recommendation can read like a sales pitch when the user did not ask for one.