Velora's AI workflows run on leading large-language models. The 14 production workflows include reply-intent routing, contact summary, email drafting, RFP response, meeting notes, COI suggestions, deal-risk scoring, content generation, and the SBC parser among others.
Every agency-aware workflow wraps its system prompt with a cachedSystem() helper that's eligible for prompt caching (5-minute ephemeral TTL). The agency profile + book context land in the cached prefix; only the per-call inputs vary. Cache stats logged on every call so hit rate is observable in production.