Olli raises $7.5 million for a family assistant betting on data protection and SOC 2 standards

The startup Olli, based in San Diego, raised $7.5 million in a seed financing round led by Kosla Ventures with participation from AI House, to fund the development of an intelligent personal assistant aimed at families that relies on text messaging and places data privacy at the core of its operating and commercial model.

The move comes as startups race to build personal agents that integrate into daily life, a space that inherently requires the assistant to access large amounts of sensitive personal data. Unlike models that rely on data collection to train the models, Olli announced that it has obtained SOC 2 compliance certification through an independent audit that demonstrates the existence of formal controls to protect customer data and operate the systems securely, making it one of the first consumer assistants aimed at families to receive this certification.

The bet here goes beyond software security to shaping a business model that ties service sustainability to paid subscriptions rather than data sales.The company's founder and CEO, Bill Lennon, who holds a PhD in artificial intelligence and previously sold his fintech platform Groundwork in 2021, explains that maintaining user trust requires not sharing their data or using it for training, which makes the monthly subscription model the only guarantee for the user’s service on their own account.

The current assistant connects to the calendar and email to organize family schedules, track tasks, plan meals and grocery shopping, book appointments and pay bills through group chats, with plans to expand into household budget management. To avoid collecting sensitive credential data, the platform does not request usernames or passwords, but carries out login and purchase actions via a private cloud browser that sends the user a remote session link to complete the transaction themselves, while awaiting the development of secure digital encoding solutions to reduce the need to re-enter data for each operation.

Competition in the text-assistant sector intensifies with tools such as Book, which was acquired by Cognition, Fam Bot, Ohai, Volk, Suner, and Tomo, alongside appointment and work-management apps like Town, Lindi, and Reclaim, as well as projects that have raised massive funding such as Instanct, which secured $350 million at a $2.5 billion valuation before launch, but faced sharp criticism for service terms that grant it perpetual, irrevocable rights to use, modify and train models on stored content.

In terms of reliability, these systems face challenges linked to the probabilistic and unstable nature of large language models. Performance issues and outages in messaging infrastructure have recurred on several experimental platforms, necessitating the construction of robust defensive frameworks around the software agent to regulate its behavior and avoid errors in service pricing or reservation execution, especially since consumer users tend to abandon inaccurate tools quickly.

This shift follows my direct work with digital product and family-finance development teams in the Gulf region, Egypt and the Levant; it shows that providing intelligent assistants that handle local correspondence and invoices requires a security architecture that isolates credential data and complies with strict local personal data protection laws. The model also demonstrates that winning users in sensitive home services no longer depends solely on context-window size, but on independent compliance attestations and the construction of isolated cloud browsers that process payments without storing users’ secrets in central databases.