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.