Unlike bots that join calls, Granola records audio at the OS level. This makes it universally compatible and positions it as a private tool, like Voice Memos, placing the onus of disclosure on the user, not the software.
While storing audio could be valuable for training models, Granola only stores transcripts. This preempts user fears of their voice data being misused or held against them, signaling a commitment to privacy over data hoarding.
Instead of tackling complex knowledge work, Granola focuses on perfecting menial tasks. This avoids the common failure mode of AI assistants that are "almost" right but ultimately useless, building user trust through consistent, reliable performance on lower-stakes jobs.
Instead of writing from scratch, the team discusses a new role on a recorded call. They then use a custom AI recipe to transform the messy transcript into a well-structured job post, ensuring it reflects the team's authentic language and collective input.
Traditional user testing creates an artificial, focused environment. Granola gets truer insights by having users share screens and walk through their actual, messy calendars and past meeting notes, grounding the conversation in reality rather than theoretical behavior.
To remove friction and encourage deep usage, Granola avoids credits or pay-per-use models, despite high backend costs. The strategy is to build the best product and capture the market first, treating inference costs as a necessary expense for growth.
The design process has shifted from comprehensive Figma mockups to live in-app prototypes built with AI coding assistants. Figma is now used sporadically to explore multiple variations of a specific component quickly, but it's no longer the start or end of the design journey.
When testing a "Coach Me" recipe that analyzed users' communication styles across meetings, the team observed powerful emotional responses. This revealed a significant opportunity for AI to provide personalized feedback and self-reflection, creating deeper value than simple productivity gains.
Granola's design philosophy targets users with "crazy work days" who are constantly context-switching. By solving for this extreme use case, inspired by OXO kitchen tools, they create a streamlined, minimalist product that benefits the average user who also experiences moments of chaos.
The founder suggests that AI systems should mimic human forgetfulness. Having an agent's memory fidelity drop off over time could be a key feature, naturally "diffusing" sensitive information from old transcripts or emails, making the system safer and more aligned with social norms.
For core product changes, Granola eschews quantitative A/B testing in favor of qualitative gut feel from intensive internal use. By building prototypes and having the entire team use them in their own chaotic workdays, decisions are made based on collective intuition about what *feels* better.
Granola's virality comes from its core utility. When a user shares high-quality, AI-generated notes moments after a call, recipients are impressed by the speed and quality, prompting them to ask, "How did you do that?" and try the product themselves.