How Revora Drafts Replies

Most routine replies run on a self-hosted model; critical replies (1–2 star, anything under your brand name) use Anthropic Claude. You approve everything.

Revora's drafting runs on a hybrid AI stack. Here's the honest version.

Most drafts: self-hosted local AI

Most routine replies run on a self-hosted model on our own hardware in Frisco, TX. This keeps costs low and responses fast.

Critical drafts: Anthropic Claude

1-star and 2-star reviews, anything posted under your business name, and enterprise/agency accounts use Anthropic Claude for maximum quality.

What shapes a draft

Revora considers:

  • Star rating — a 1-star reply reads very differently from a 5-star one
  • Your tone profile — on Pro and Multi-Location, Revora learns your voice from your edits
  • Business type — a dentist's reply differs from a gym's
  • Review content — specific concerns get specific acknowledgment

The model is plumbing

We treat the underlying engine as implementation detail. What matters is whether the output is good, on-brand, and ready for your approval.

Spanish drafts (Pro and Multi-Location)

Pro and Multi-Location support Spanish-language drafts for tone-tuned replies. Other languages are case-by-case — email us if you need one and we'll tell you honestly whether the output meets our quality bar yet.

Was this article helpful?
Still need help?
Ask Reva right here, or email our team — a real person replies.