TORONTO
Harmix has raised $1 million in funding to support the deployment of Pam (Proactive AI Manager). The platform is designed to help small and medium-sized businesses unify data across fragmented software tools.
The Self-Onboarding Problem
Even tools with thousands of integrations today require users to manually share their preferences: typing context into a chat session (e.g., "I prefer formal tone with investors"), or writing and maintaining a preferences file (e.g., SOUL.md in OpenClaw, where users document their working style, relationships, and priorities). In theory, this gives the agent a user model. In practice, most users do not know what to write, and fewer still keep the file updated as their context changes.
The result: the agent performs well on generic tasks (e.g., researching news online, summarizing a public document) but fails on anything contextually nuanced (e.g., drafting a follow-up to a client with two years of relationship history, writing an RFP tailored to what this specific customer actually cares about, or auto-replying to emails with the right tone for each person).
What Pam Does Differently
Rather than asking users to describe themselves, Pam has what we call self-onboarding memory. It reads the organizational context users already live in: their email threads, Slack messages, calendar meetings, and AI chats. From this raw data, it builds a working model of who they are, how they communicate with each specific person, and what happened in each situation.
Critically, Pam does not just passively read raw data. It identifies conflicts (e.g., a contact's role changed but their record was not updated; a deadline in email does not match what is in the calendar) and asks the user proactive questions to fill gaps that are not documented anywhere. This means Pam's context stays current without any manual effort, and the outputs it produces reflect real preferences, real relationships, and real priorities.
This self-onboarding memory is also the foundation for the productivity workflows we are building on top of it.
"Most AI tools are reactive and require a specific user prompt to provide value. For smaller businesses where resources are limited, managers cannot always spend time asking the right questions. Pam is engineered to monitor these fragmented interfaces proactively to identify bottlenecks and track decisions."
— Nazar Ponochevnyi, CEO of Harmix
Use of Funds
The new capital will be used to accelerate the delivery of Harmix's 2026 technical roadmap, with a focus on data sovereignty and security. Funds will be allocated to:
- SOC 2 Certification: Obtaining industry-standard security compliance
- Self-hosted Deployment Options: Developing solutions that allow businesses to maintain full control over their proprietary data on private clouds
- Engineering Team Expansion: Scaling operations within the North American and European manufacturing sectors
About Harmix
Harmix is an AI company that built a multimodal search API for media brands, helping teams search and manage more than 30 million music and video assets. While building a profitable business, the team also used AI internally for lead generation, staff training, dashboards, and repetitive workflows, but found existing tools too limited and disconnected from how real companies operate. That led Harmix to build Pam, a proactive AI manager that connects knowledge across emails, meetings, chats, documents, and ERPs to coordinate teams, automate routine work, flag issues early, and uncover growth opportunities.