GeraVoice vs. Twilio AutoPilot vs. Voiceflow vs. USSD: Adjacent Projects Compared
Published 21 April 2026 · 9 min read
Twilio AutoPilot / Voice Intelligence
Twilio is the de facto telephony primitive for developers. Their voice AI stack handles ASR, TTS, and intent routing well, and their global number infrastructure is unmatched.
Overlap: telephony + ASR + TTS + intent.
Gap: no published intent schema across services, no opinionated confirmation flow, no enforced human fallback SLA, no minimisation-first data access to user vaults.
Relationship: complementary. GeraVoice can sit on top of Twilio for call routing and ASR — our work is the layer above.
Voiceflow
Voiceflow is a strong design platform for conversational applications. Drag-and-drop flow, version control, collaborative design. Excellent for voice-app authors.
Overlap: conversation design.
Gap: app-builder surface, not a protocol. Built for call centres and Alexa skills, not for cross- product service gateways. Limited emerging-markets presence and no opinion on the 2G-and-feature-phone constraint set.
Relationship: complementary for teams who want a visual design surface feeding into a GeraVoice deployment.
USSD menus
Unstructured Supplementary Service Data is the long-standing feature-phone interface (dial *123#, navigate by digit). M-Pesa and large swathes of mobile money run on USSD. It is reliable, ubiquitous, and cheap.
Overlap: universal reach to feature phones.
Gap: text-only (in a short 182-char line), no natural language, rigid menu trees, language-by-network limitations, poor accessibility for non-literate users.
Relationship: complementary. A USSD menu that connects into GeraVoice for the complex paths is a reasonable hybrid — USSD for discoverability, voice for depth.
Open-source voice-agent frameworks (Vocode, LiveKit Agents)
Vocode and LiveKit Agents are developer-friendly toolkits for building voice AI agents. Both are under active development and well-architected.
Overlap: voice-agent runtime.
Gap: runtime, not protocol. Teams using these still have to invent their own intent schema, confirmation flow, consent model, and audit log.
Relationship: GeraVoice protocol spec could be hosted inside a Vocode- or LiveKit-built runtime.
Where we are genuinely different
- Opinionated intent schema across verticals. Bookings, payments, queries share a schema so agents and humans can handoff without translation.
- Numeric confirmation mandatory above the micro- tier. Not a suggestion, a spec requirement.
- Human fallback SLA. Published, monitored, first-class in the protocol.
- Integration with GeraMind. Scoped queries for user preferences, consent, identity.
Where we might be wrong
- Opinionation can be a bug. Teams with unusual flows may find the schema constraining; we will add extension points but keep the core tight.
- Human fallback is operationally expensive. If it cannot scale below certain transaction margins, the product is narrower than we hope.
- Language coverage depends on continued ASR investment in low-resource languages. That is outside our direct control.
Cooperative future
The right stack for 2030: a telephony primitive (Twilio or equivalent), an ASR/TTS layer (vendor-agnostic), a voice- agent runtime (LiveKit, Vocode, or similar), GeraVoice protocol on top, cross-product integration with GeraClinic, GeraEats, GeraCash. Each layer independent, each swappable. We want to be the right answer in the protocol layer.
Help build voice-first commerce.
Join the waitlist