Content Quality & Brain
How HarperFlow learns your brand's voice during onboarding, and how to change it any time afterward.
HarperFlow writes in your brand's voice from the very first article, built from what you told it during onboarding. You're never locked into that original answer.
Your profile includes a voice and tone, a brand archetype, your audience, your positioning, and a one-line description of what you do. HarperFlow derives an initial draft of this from your onboarding answers, and its writer references all of it on every article it produces afterward, from headline phrasing to how directly it talks about your product. A profile might describe a direct, no-nonsense tone aimed at technical buyers, or a warmer, more encouraging one aimed at first-time users; articles read noticeably differently depending on which you've set.
Your brand voice lives on your project's settings and is reflected back to you on your Org Brain page. Editing any field there applies to articles written from that point forward. It doesn't retroactively rewrite anything already published; past articles keep the voice they were originally written in. You don't need to touch every field at once either; changing just the one that's off, your positioning, say, is enough to shift future articles.
A change here doesn't go through some separate re-training step. The very next article HarperFlow drafts, whether from autopilot or an on-demand run, reads the updated profile directly, so you can see the effect of an edit in the next article that comes out rather than waiting on a longer cycle.
Note: If your organization has shared memory turned on across multiple sites, the shared org-level voice can take precedence over what's saved on an individual project. See Org Brain and shared memory for how that resolves.