Instinct’s text-message AI agent is holding its own as OpenAI and Meta close in

Instinct, the invite-only AI assistant startup that emerged in August with almost no marketing, is still performing well even as bigger rivals from OpenAI and Meta enter the same consumer-agent market. The company’s text-message-based assistant has remained competitive in hands-on testing, but its momentum now faces a major question: whether a startup can stay buzzy once tech giants bring similar products to market.

That matters because AI agents are quickly becoming one of the most contested corners of consumer AI, and Instinct is trying to prove that a focused product can beat larger platforms built by companies with far deeper resources.

Before the current wave of cartoonish AI companions and branded assistants, Instinct was already winning attention the old-fashioned way: by being useful. The startup launched quietly, without a public app or splashy campaign, yet word spread quickly among early testers who praised its simple texting interface and its ability to handle real-life chores such as scheduling appointments, organizing follow-ups, and navigating everyday admin work.

Now the competitive landscape is much more crowded. OpenAI has rolled out Dot, and Meta has introduced Muse, both of which are aimed at similar consumer tasks. But after testing Instinct alongside those products, the startup still appears capable of keeping pace in practical usage, even if the technology remains constrained by the messy reality of the web and the limitations of current AI agents.

What makes Instinct different from the new AI rivals?

Instinct stands out because it behaves less like an app and more like a quiet assistant that lives inside your messages.

Unlike the mascot-driven experiences being pushed by some larger competitors, Instinct uses a stripped-down approach: conversations happen through text messages, email, iMessage, WhatsApp, or similar channels. There is no standalone consumer app for browsing through features, and there is no cute avatar sitting onscreen pretending to work on your behalf. Instead, the assistant sends short replies, occasional emoji reactions, and practical updates while handling tasks in the background.

That design choice is central to the company’s pitch. Founder Noah Shinn says Instinct is trying to create a personal assistant for everyday life rather than a general-purpose AI platform. In his view, the product’s lack of a traditional interface is not a weakness but a differentiator. It is meant to fit into the places people already communicate, whether that is text, email, or another messaging tool.

Shinn says the company is focused on building an assistant that “just gets it,” arguing that Instinct is designed to turn conversation into action rather than merely answer prompts.

The broader market is moving in a different direction. OpenAI and Meta are building AI systems across multiple use cases and audience segments, while Instinct is deliberately narrower. That narrower focus may be why the startup has been able to build a product that feels more immediately practical for people who want help with daily life rather than a generalized chatbot experience.

How does Instinct actually work in daily life?

Instinct works by combining conversation with access to outside services, including a virtual machine and selected personal accounts.

Users can connect the assistant to tools such as Google Workspace, Slack, and Notion, and when the system needs login credentials or a website-specific action, it directs users to a secure vault link. The result is a lightweight interface that mostly disappears into the background while the assistant does the work.

That approach gives Instinct a practical advantage: it can be used anywhere you can send or receive texts. In theory, that makes it compatible with settings that are awkward for full apps, such as CarPlay or smartwatches. The experience is meant to feel available without demanding much attention.

There is also a privacy-and-comfort tradeoff. Having a helper embedded in message threads can feel convenient, but it can also feel unusually close. The assistant’s always-nearby nature may be ideal for some users and unsettling for others, especially because it is acting on personal data and personal schedules.

The daily tasks Instinct is built for

Instinct is aimed squarely at life-admin work, not abstract brainstorming. In testing, it handled tasks such as finding swim lessons, emailing a doctor’s office, and beginning an Ikea return. These are the sort of chores that are easy to postpone but annoying to do by hand.

That makes the assistant part of a larger shift in AI products: the move from conversation to execution. Instead of answering questions and stopping there, the system attempts to complete tasks that require a chain of actions across websites, forms, and emails.

• Searching for appointments and service options
• Drafting and sending follow-up emails
• Filling out forms and return paperwork
• Pulling together relevant account or calendar details
• Flagging possible conflicts or missing information

Why is this moment so important for Instinct?

This moment matters because Instinct’s early advantage was built on scarcity, and scarcity is harder to preserve once major companies copy the format.

The startup became prominent in part because it moved first with a consumer assistant that felt polished, useful, and different from the noisy world of generic chatbots. It also benefited from a low-profile launch strategy: invitation-only access, no app store presence, and barely any public-facing marketing. That controlled rollout helped fuel interest and gave the product an air of exclusivity.

That strategy has also drawn unusual investor confidence. By late September, even after competitors had arrived, Instinct was valued at $10 billion, a sign that backers believe the company still has room to grow despite the new pressure from larger rivals. For a startup with no monthly fee and no public app, that valuation suggests investors are betting on product quality, future monetization, or both.

Shinn’s background may also have helped the company gain credibility. He previously worked as an early employee at Sierra, another highly valued startup known for its Silicon Valley pedigree. That kind of background matters in a market where trust, technical ambition, and investor confidence are all tightly linked.

How did the assistant perform in hands-on testing?

Instinct performed well enough in testing to look genuinely promising, though not flawless.

It initially stumbled over basic context, such as assuming a Cincinnati area code meant the user still lived there, but it adapted once corrected. After that early misfire, the assistant handled a steady stream of practical tasks better than expected, often producing helpful results without requiring overly specific instructions.

One of its strengths is that it can cope with vague prompts. When asked to search for an Ikea product described only as “vertical and white” with branch-like parts, it figured out the user meant the Tjusig coat rack. That is a small example, but it shows a useful trait: the assistant can infer intent rather than forcing users to know exact product names or technical details.

Where Instinct impressed the most

Instinct’s most impressive moments came when it went beyond basic automation and offered useful context on its own.

For example, while helping schedule an eye-doctor appointment, it identified the likely office even though the exact name had not been remembered at first. It later surfaced a confirmation email from 2024 to reassure the user it was working with the right office. In another case, it noticed a calendar conflict while completing a separate task.

That kind of behavior may sound modest, but it points to a more meaningful use case than simple autofill. Filling in forms is helpful. Recognizing relevant details, connecting the dots, and surfacing important information before the user asks for it is much closer to what people actually expect from a real assistant.

In testing, the assistant proved strongest when it could identify useful information on its own instead of waiting for exact instructions.

What are the limits of today’s AI agents?

Today’s AI agents are still only as good as the websites, forms, and data they encounter, and those systems are often messy.

The swim-lesson example shows the problem clearly. Instinct found age-appropriate class options and correctly narrowed suggestions to a community center when prompted. It also matched dates and times accurately, which is a basic but necessary accomplishment. However, it missed a crucial prerequisite class requirement hidden elsewhere on the site. The information was there, but not in an easy-to-process place.

That is a common failure mode for agentic AI. Unlike clean database tasks, the open web is full of pop-ups, buried notes, contradictory information, and layout quirks created by humans over many years. A system can appear intelligent and still miss the one detail that matters most.

In practice, that means AI agents are good at reducing effort, but they are not yet reliable enough to replace human oversight for important decisions. They can search, draft, click, and submit. They cannot always judge whether what they found is complete or whether a hidden rule changes the outcome.

Who is Noah Shinn, and what is his strategy?

Noah Shinn is the founder of Instinct and the architect of its narrow, utility-first strategy.

His background includes early work at Sierra, which helped establish his credibility in the startup world before he launched Instinct on his own. At Instinct, the strategy has been to focus intensely on one thing: a personal assistant that helps with the practical details of everyday life.

That emphasis on specificity may be why the company has managed to stand out. While large AI platforms are trying to cover many categories at once, Instinct has centered itself on personal chores, scheduling, follow-ups, and small but time-consuming administrative tasks. The product is not trying to be everything. It is trying to be useful enough that people actually return to it.

Shinn is also signaling patience on monetization. Asked whether the service will stay free, he framed the current free access as a phase tied to invite-only growth rather than a permanent promise. That is typical of venture-backed startups, which often prioritize usage, retention, and market position before they decide exactly how to make money.

Why are analysts watching the business model so closely?

Analysts are watching closely because consumer AI agents may be expensive to run, difficult to monetize directly, and attractive to advertisers and platform owners for different reasons.

Avi Greengart of Techsponential says startups like Instinct usually need growth first, because growth is what makes them valuable. The near-term objective is not necessarily to maximize profit but to build something big enough to justify future payment, whether through subscriptions, enterprise deals, acquisition interest, or another business model.

There is a separate question of who ultimately pays if a service seems free. Greengart argues that if users are not paying directly, the costs may eventually be recovered in another way, possibly through partnerships or commerce relationships that make the assistant more commercially oriented.

Greg Ireland of IDC makes a similar point, noting that consumer assistants that engage in shopping or transactions naturally lend themselves to advertising. He argues that these products may eventually become more ad-driven as the market matures, especially if they help users buy products, book services, or make purchases inside the assistant flow.

Analysts suggest that if consumer AI agents are not funded through subscriptions, their economics may push them toward advertising, commerce referrals, or other indirect revenue sources.

What does the competition mean for the broader AI market?

The competition around Instinct reflects a bigger turning point in consumer AI: the move from demo-worthy tools to products that are supposed to actually do things for people.

That shift has drawn in the biggest names in tech because AI agents sit at the intersection of messaging, productivity, shopping, search, and personal data. Whoever wins users’ trust in that environment may gain a powerful foothold in how people interact with software every day.

OpenAI’s Dot and Meta’s Muse are not just copycat releases. They represent the scale advantage that large companies bring to fast-moving product categories: distribution, compute, existing user bases, and the ability to iterate rapidly. That raises the pressure on startups like Instinct to prove they offer something more than novelty.

Still, being first to market can matter. Instinct’s early reputation was built on being lean, practical, and focused, and those qualities may help it retain loyal users even if the market becomes crowded. The question is whether the product can keep improving fast enough to stay ahead of companies with much bigger teams.

What comes next for Instinct?

What comes next for Instinct is a test of whether usefulness can remain more persuasive than scale.

If the product continues to handle real-world chores smoothly, it could carve out a durable niche as a text-first assistant for busy people who do not want to navigate another standalone app. If larger players replicate its strengths while bundling them into larger ecosystems, Instinct may have to rely on speed, specialization, and trust to preserve its edge.

For now, the startup remains in a strong position relative to its size. It still has a compelling product, a strong valuation, and a format that feels distinct from the cute-persona trend embraced by its rivals. But the arrival of OpenAI and Meta changes the stakes. Instinct is no longer the only new idea in town. It is one contender among several, and the race to define what a consumer AI agent should be has only just begun.

Key facts at a glance

Timeline: how the contest escalated