OLISE 2026 roadmap: real barge-in, pt-BR, SOC 2 and more
What we are going to build, in what order, and which milestones will keep us up at night. No marketing.
OLISE Team
Product
Publishing the roadmap is a commitment. We do it because we want our customers to understand where we are focused and why.
Q2 2026
Real barge-in (Twilio Media Streams + VAD). Today our interruption detection is good but not perfect: there is a small window where the customer starts talking and the AI is slow to stop. We're targeting <100ms TTS close on detected human speech.
Native Brazilian Portuguese. Voice, model, and tone tuned for Brazil. Our Florida customers have a high share of Spanish speakers, and also Brazilians. It is the natural next language.
SOC 2 Type I audit. In progress. Type II ramp set for H2.
Q3 2026
Unified inbox (mail + voice + web chat) with federated search. Consolidates what we already have.
Pro reports: ticket MTTR, automatic clustering of complaints, alerts. High-volume customers need this now.
Public API v1 for custom integrations. Today it is internal; we want to open it with proper docs.
Q4 2026
European market: EU infra, data residency, improved DPA, EUR billing.
Custom voices: for Enterprise customers who want to clone a brand-aligned receptionist voice. With explicit, signed consent.
What we won't do
- We won't build a generic web chat or "assistant for everything." B2B premium phone operator is the lane.
- We won't sell call data to train models. Contractual policy.
- We won't accept verticals outside services (no healthcare unless private clinic with consent, no regulated finance).
How you can help
If you have a use case you think fits and isn't covered, write to us. If you've been with us for six months and think something above is mis-prioritized, write too. The roadmap is a living document.
blog.related
blog.category.engineering
Why we picked Anthropic Claude for our AI phone operator
We compared Claude, GPT-4o, and Gemini on natural Spanish, latency, and real tool use. When voice is the channel, the gap shows.
blog.category.engineering
How we built prompt-injection defense in production
Three layers of defense, an output filter with regex, and why we never trust a single line of protection. Real case with code.
blog.category.industry
Case study: how a Founder Pricing partner scaled their phone coverage
A car rental operation in Florida. Missed calls before and after OLISE. Lessons from the first 90 days.