Hello Haven Raises $15M for Personal AI Digital Twins

Hello Haven has raised a $15M pre-seed round led by Mayfield to build what it calls a personal AI digital twin. The company emerged from stealth on September 17 with Haven, a product designed to remember a person's preferences, relationships, history, and decisions, then use that context to complete work across digital channels.

The financing is an early wager on a specific problem in consumer AI: an assistant can produce a sophisticated answer and still know almost nothing about the person asking. Hello Haven's response is a private, persistent memory graph that is meant to travel across models, tools, and interfaces instead of resetting whenever the user changes apps.

A $15M Bet on Persistent Personal Context

Frank Addante, Hello Haven's founder and CEO, and co-founder Duc Chau previously helped build The Rubicon Project. That history matters because both businesses depend on orchestration. Rubicon's infrastructure routed complexity behind advertising transactions; Haven must decide which model, tool, permission, and moment should serve an individual request.

Mayfield managing partner Navin Chaddha said his firm backed Addante at Rubicon Project decades ago and is backing him again. MVP Ventures separately identified Hello Haven as a portfolio company and said it invested, although Hello Haven's release did not disclose the full syndicate, valuation, or financing terms.

The round is described as pre-seed, which makes its size notable but not self-validating. It gives the company time and resources to build infrastructure, integrations, and distribution. It does not by itself establish product-market fit, defensibility, or paid demand.

What Haven Says It Can Do

According to the company's announcement, Haven works across phone, text, email, voice, web, and mobile. The company lists daily briefings, inbox triage, follow-ups, returns, orders, and calls among the jobs the system can handle, including tasks that recur without a fresh prompt.

Hello Haven says its orchestration layer can choose among more than 500 AI models and technologies and connect to more than 100 applications. It also reports more than 2M interactions since opening private beta one month before the financing announcement. The product is available on the web, Apple's App Store, and Google Play.

Those metrics require careful interpretation. Interactions are not the same as users, retained users, paying customers, or successfully completed tasks. Hello Haven did not disclose customer count, revenue, paid conversion, retention, or an independent performance benchmark in the materials reviewed for this article.

Memory Is Both the Product and the Risk Surface

The strategic idea is that models will keep changing, while personal context can become the durable layer. If Haven remembers which details matter, understands a user's relationships, and can select among different tools, the user should spend less time restating intent. A future developer platform could extend that context into apps, televisions, vehicles, connected devices, and robotics.

That same architecture raises the standard for trust. NIST's AI Risk Management Framework emphasizes managing AI risks throughout design, deployment, and use, while its work on agent standards highlights security and interoperability as systems act across tools. OWASP's guidance for AI agents recommends least privilege, memory isolation, input validation, and human approval for high-risk actions.

Hello Haven says a person's information is stored in a private memory graph that the individual owns and controls. That is an important product promise, but the reviewed sources did not provide an independent security audit, certification, detailed permission model, or public red-team results. As the product gains the ability to act, evidence about reversibility, access controls, data portability, and memory integrity will matter as much as model quality.

What the Capital Makes Possible

Hello Haven's careers page lists roles in integrations, voice, optimization, robotics integration, quality assurance, strategic partnerships, operations, and marketing. Those openings line up with a product that must combine model selection, application connections, natural interfaces, and reliable execution rather than win on one model alone.

The company has not assigned the $15M to precise budgets, hiring targets, geographies, or milestones. Its stated roadmap points toward a developer platform and broader device coverage, but execution will depend on whether integrations are dependable enough for daily use and permissions remain understandable as the system becomes more autonomous.

The Next Proof Point Is Earned Authority

Hello Haven is entering a crowded field of assistants, agent platforms, and device-level AI. Its clearest differentiation is not a claim to have the single best model. It is the attempt to make a person's memory, preferences, and decision history portable across whatever model or interface is best for the moment.

The $15M round gives Addante, Chau, and their team a chance to prove that continuity can become a product rather than a feature. The important measures from here will be retained use, completed-task quality, permission safety, paid demand, and independent evidence that the memory layer remains under the user's control. If personal AI becomes valuable by knowing more, Hello Haven's challenge is to earn the authority to act without asking users to surrender the very control the product promises to protect.