Fundraise in this Series C: $1 billion (Source: TechCrunch (September 28, 2026))
Delegation is a permission model, whether or not anyone drew it
Hand an assistant your credit card, your calendar and your phone number, and you have made a set of trust decisions. Hand the same access to software and you have made the same decisions; what changes is where they get written down. Instinct's round is a useful case for asking what an agent is given, and on what terms. TechCrunch's concern was narrower than permissions: some users question how much personal information they must disclose, and it called the company's first privacy policy worrisome for its overreach. A privacy policy governs data collection, so we read it as the data-side half of the permission question. What an agent may do on a user's behalf would sit in other documents too, such as terms of service and in-product settings, which the source does not discuss. That framing is our interpretation, and it rests on a single company.
TechCrunch reports that the startup pulled in $1 billion, with Sequoia Capital, Benchmark Capital and Coatue among the investors. The sum sets its valuation at $10 billion, up from the $2.5 billion it announced a month before. The Information reported the raise earlier in September, and a Monday press release from Instinct confirmed it as a Series C. The service itself only began admitting users, by invitation, in August 2026.
Reading a question is a read. Paying a bill is a write.
TechCrunch describes a class of consumer agents that goes past conversation: booking travel and restaurant tables, buying things, paying bills, cancelling subscriptions, ordering groceries. When Instinct takes on a task, the article says, it works from a phone number and a computer of its own, and it talks to users over SMS. Recent additions include a concierge feature that phones venues without online booking, and a "trusted person network" in which one person's agent coordinates plans with friends' agents.
The difference matters to a budget-holder. An assistant that answers questions reads information. One that moves money, ends contracts and speaks to third parties writes to the world on someone's behalf. Whatever access an agent is granted to do those things, the rules around that access shape the outcome.
Some users are questioning how much they must disclose
TechCrunch reports that some users question how much personal information they must give agents to unlock these features. It says Instinct's first privacy policy was especially worrying because of its overreach, and that the policy has since been updated. The article does not say what the original text contained, what changed, or why, so we do not characterize any of that. The narrow point that the source supports is this: a policy TechCrunch described as overreaching was in place first, and it was later revised.
Instinct has not shared user numbers or growth metrics, TechCrunch notes, though "clearly its investors are seeing something they like." Meta's Muse, by the same account, offers many similar features plus links into Meta's social products, such as summarizing Instagram DMs or Facebook Groups, and has reached the top of the U.S. app stores with millions of downloads. Instinct has no mobile app yet.
Who carries the risk
In our reading, the person delegating carries the first layer: their data, payments and reputation travel with the agent. The second layer sits with the organizations on the other end. Per TechCrunch's description of the concierge feature, a business that takes bookings by phone can now receive a call from an agent acting for a customer. That is a design goal of the product, not a fault. It does mean a front desk, a support line or a retention flow is dealing with a counterpart it cannot assume is human.
This is one company and one funding round. It shows investor appetite, not measured adoption, and Instinct has disclosed no usage figures. We treat it as a signal about what is being built, not a count of who is using it.
Questions to put to your team
First, do our acceptable-use rules address staff giving consumer agents access to work email, calendars or payment methods? Second, how do our phone lines, support channels and cancellation flows verify that a request from an agent is authorized by the customer it names? Third, when we evaluate any agent product, who reads the privacy policy, and do we track its version history and retention terms? Fourth, which of our customer processes assume a human is on the other end, and what happens when one is not?
Our view: permissions are easier to define before the first task than after the first headline.
Produced by the WebPulse Newsroom with AI assistance from the original reporting credited below, and checked against that source by our editorial review. How we use AI.
Original reporting: TechCrunch.
