Building Earmark: How a Two-Person Team Turned Meetings Into Finished Work

Key Takeaways

• Real‑time AI agents generate specs, tickets, slides during calls
• Ephemeral architecture satisfies enterprise data‑privacy requirements
• Prompt caching reduces meeting AI cost below $1
• Pivot from Vision Pro coaching to web meeting assistant broadened market
• Agentic search combines RAG, BM25, metadata for deeper insights

Summary

Earmark’s new productivity suite transforms live meetings into concrete deliverables such as product specs, tickets, and slide decks, eliminating the traditional post‑call cleanup. The platform runs several specialized AI agents in parallel, translating jargon, drafting documents, and even spawning prototypes while participants speak. After an early pivot from an Apple Vision Pro coaching tool, the company embraced a web‑based, ephemeral architecture that appeals to enterprise customers concerned about data residency. By leveraging prompt caching, Earmank slashed its per‑meeting AI spend from roughly $70 to under a dollar.

Pulse Analysis

The rise of AI‑driven meeting assistants promises to streamline collaboration, but most solutions stop at transcription or generic summaries that quickly gather dust. Earmark differentiates itself by delivering actionable artifacts in real time, a capability that directly addresses the hidden labor cost of turning discussions into implementation tasks. This approach resonates with product teams that juggle rapid iteration cycles and need immediate, reliable documentation without manual follow‑up.

Under the hood, Earmark orchestrates a fleet of purpose‑built agents that run concurrently during a call. One agent translates technical jargon, another drafts specification outlines, while a third spins up prototype code in tools like Cursor or V0. The company’s prompt‑caching layer stores reusable LLM prompts, driving per‑meeting AI expenses from a prohibitive $70 down to under a dollar. An "ephemeral" mode ensures no conversation data is persisted, turning a privacy concern into a marketable feature for enterprises bound by strict data‑governance policies.

For the broader market, Earmark’s model signals a shift toward AI as a true chief of staff rather than a passive recorder. By targeting product managers as extreme users, the platform validates a high‑value use case that can be extended to other functions such as legal review or security compliance via persona‑based agents. As organizations seek to embed AI deeper into workflow automation, solutions that combine cost efficiency, privacy, and tangible output—like Earmark—are poised to become indispensable components of the modern digital workplace.