Instinct built its reputation as an AI assistant that can take on everyday online chores through a conversation by text. After OpenAI’s Muse and Meta’s Dots entered the field, the startup faced a new question: could its focused approach hold up against products from larger companies? In testing, Instinct remained useful, though its mistakes showed why human oversight still matters.
A personal assistant in your messages
Instinct arrived with little marketing and an invite-only launch. Its straightforward, text-based interface and ability to handle tasks such as booking appointments and sending follow-up emails quickly attracted attention. The service has no app or monthly subscription fee for now, and access requires an invitation from a current user or approval from a waitlist.
Rather than asking users to open a dedicated app, Instinct works through iMessage, WhatsApp, or email. It can connect to a virtual machine and, with permission, to services such as Google Workspace, Slack, and Notion. For website logins, it sends a link to enter account details in a “secure vault.”
That design makes the assistant available wherever users can send and receive messages, including on a watch or through CarPlay. It also means there is little ceremony: users describe a task, and Instinct responds briefly while doing more of the work in the background.
Founder Noah Shinn says the company’s focus on personal assistance distinguishes it from broader AI platforms. “We have been solely focused on building a personal assistant for everyday life that just gets it.”
Useful with messy, everyday requests
Instinct was tested on the kind of administrative tasks that accumulate in a busy household. It found swimming lessons suited to a child’s age, helped start an Ikea return, and emailed an eye doctor’s office to arrange an appointment. The assistant could also interpret an imprecise description of an Ikea purchase and identify the item from past orders.
Some of its most helpful moments came when it found context the user had not explicitly requested. While helping contact the doctor, it located a confirmation email from 2024 to help verify the office. During a separate chore, it spotted a possible calendar conflict.
Those examples suggest that an AI agent can do more than fill in forms. It may connect a vague request with information in a user’s email or calendar and surface a detail that helps move the task forward. That can save effort, provided the information it finds is relevant and correct.
People still need to check the details
The swimming lesson search also exposed a meaningful limitation. Instinct found workable options and returned times and dates that checked out when reviewed in a browser. But it missed a prerequisite class required for children enrolling in preschool-age lessons.
The omission mattered because the information was not simply a date or a form field. It was a condition that changed whether the suggested lessons were suitable. A detail buried in a pop-up or elsewhere on a human-made website can be easy for an automated agent to overlook, even when the rest of its results look convincing.
That is why checking an agent’s work remains part of using it. Instinct can gather options and handle routine steps, but a user may still need to confirm requirements, review submissions, and decide whether the result fits the situation.
A different approach in a crowded field
Muse and Dots offer similar help with online tasks, but present themselves with mascots and more visible consumer-facing personalities. Instinct takes a quieter route: a messaging conversation, brief replies, and actions carried out on the user’s behalf.
That simplicity can make an assistant feel conveniently close at hand, while also raising questions about how much access it should have to personal services. Instinct’s model depends on connecting accounts and letting software interact with websites, email, and calendars. Its usefulness therefore rests on both task performance and the user’s comfort with that access.
Instinct’s focused product has so far shown it can compete on practical chores, even as larger companies enter the category. Its strongest case is not that every task runs perfectly, but that it can turn ordinary messages into useful action. The missed swimming prerequisite is a reminder that, for now, the person asking still has an important role: checking that the assistant has understood the whole task.