Instinct raised $350 million before it even launched, and still managed to generate bad press for a privacy policy that granted itself a “perpetual and irrevocable” license to use, store, and train on anything users shared. That’s the kind of opening a smarter competitor should exploit. According to TechCrunch, San Diego-based Ollie is trying to do exactly that.
Ollie is a text-based personal assistant built around family life. It connects to your calendar and email, helps organize schedules, plans meals, tracks to-dos, shops for groceries, books appointments, and handles bill payments through group chats. Budget management is on the roadmap. That’s a wide surface area of personal data, which makes Ollie’s core argument worth examining: that it can handle all of this without becoming a liability for the people using it.
The company has achieved SOC 2 compliance, which makes it one of the first mainstream, consumer-facing AI assistants to do so. SOC 2 is an independent audit framework that verifies a company has formal controls in place to protect customer data. It’s a standard more common in enterprise software than in consumer AI, and its presence here is a deliberate signal. Ollie runs on a subscription model, meaning the business isn’t built on monetizing user data. “We’re not sharing your data with anyone,” said co-founder and CEO Bill Lennon. “Ollie works for you.”
Lennon has a PhD in AI and sold his previous startup, Groundwork, a neobank for nonprofits, in 2021. That fintech background matters as Ollie pushes further into household money management. The company has raised a $7.5 million seed round, primarily from Khosla Ventures, with participation from AI House. That’s a fraction of what Instinct pulled in, but Ollie isn’t trying to win on scale right now.
The competitive field is crowded. Text-based consumer assistants include Poke (acquired by Cognition), Fambot, Ohai, Folk, Saner.ai, and Tomo. Work-focused tools like Lindy, Town, and Reclaim.ai occupy adjacent territory. And Instinct, despite its PR stumble, still has enormous capital behind it. Ollie’s privacy positioning is a real differentiator, but differentiation alone doesn’t win markets.
One technical decision worth noting: Ollie does not ask for usernames or passwords. When it needs to log into a site on a user’s behalf, it uses its own cloud browser and sends the user a link to complete the session remotely. The same approach applies to payments. It’s more secure than storing credentials, but it also means users have to manually complete certain steps every time. Lennon acknowledged this and said tokenization is on the roadmap, framing it as a balance between convenience and trust that the technology isn’t quite ready to solve cleanly yet.
Reliability is the harder problem. During testing, Ollie went offline due to an infrastructure issue with its text provider. Instinct, in a separate test, returned incorrect hotel pricing. These aren’t isolated bugs. They reflect a broader issue with building dependable agent systems on top of probabilistic language models. Lennon described it directly: “They’re stochastic, they’re inherently unreliable. We have to build the agent harness in a defensive way to catch and prevent those things. It’s almost like there are just 1,000 cuts you’ve got to solve first.”
That’s an honest assessment, but it also frames the real challenge for Ollie and every assistant in this space. Users tend to give new tools one shot. If it breaks early, they leave. Retention data Lennon cited is encouraging, he says the curves are consistent with leading AI subscriptions, though he declined to share user numbers, citing competitive sensitivity.
The privacy angle is the right bet at the right time. Consumer trust in AI products is fragile, and broad data policies are starting to generate real backlash. But Ollie still has to prove it can be both trustworthy and reliable. Those two things together are what the market actually needs.
