TeleVoice — AI Voice Calls via Telegram

The company appears to be a solo project (1 person) named TeleVoice, built as a hackathon submission for OpenAI Build Week 2026. The author states it is an AI voice calling system that integrates Telegram, GPT-5.6, and Twilio to enable autonomous sales calls in Spanish or English. It was built using Codex and OpenAI tools over four days.

What changed: The project evolved from a proof-of-concept into a potential standalone product (TeleVoice) that may be integrated into the author's main business, Dopa, an Agentic ERP for SMBs in Latin America. The author claims this demo is a version of voice technology already powering Dopa’s offerings.

The single most important open question: Is there any evidence of traction, revenue, or customer adoption beyond the author’s own demonstration and stated intent?

This analysis is based solely on the self-reported project description provided by the caller. No external verification, archived data, or third-party sources are available. All claims in this report are labeled as “the description states” and should be treated as unverified.

• The description states that TeleVoice is a Telegram bot that triggers AI voice calls via Twilio, using GPT-5.6.
• It allows users to message the bot with a request like: “Call Juan at +51 979 300 062, he wants to buy Dopa.”
• GPT-5.6 extracts phone number and context, Twilio places a real PSTN call, and an AI voice agent (named Sofia) handles the conversation in Spanish.
• The system is built using Node.js, Express, TypeScript, and integrates with Telegram Bot API, Twilio Voice, OpenAI SDK, and Codex.

Not evidenced: No actual product or live deployment exists beyond the author’s own demonstration. No customer data, usage metrics, or real-world call logs are provided.

• The description states that TeleVoice is a “Telegram bot + GPT-5.6 + Twilio = autonomous AI sales calls in Spanish or English.”
• It was built to address the problem of lost sales due to delayed responses in small businesses, particularly in Latin America.
• The author positions it as an extension of their existing product, Dopa (Agentic ERP), and claims that this demo is a standalone version of voice tech already powering Dopa.
• It is described as a “voice layer” to complement text-based AI agents already used in Dopa.

Inferred: The positioning suggests TeleVoice is intended for B2B SaaS sales automation, but no evidence supports actual customer use or market traction.

• The description states that the author runs Dopa, an Agentic ERP for SMBs in Latin America.
• The target customer is likely small to medium-sized businesses (SMBs) in Latin America who lose sales due to delayed responses.
• The AI voice agent is designed to close deals via phone calls, suggesting a sales enablement use case.

Not evidenced: No explicit customer list, persona data, or segmentation details are provided. The ICP is inferred from the context of Dopa’s business model.

• The description states that TeleVoice is part of a larger product ecosystem — Dopa (dopa.solutions).
• It is described as a standalone version of voice technology already powering Dopa, suggesting it may be integrated into a broader platform or workflow.
• No pricing model, revenue streams, or monetization strategy are mentioned.

Not evidenced: No information on how TeleVoice would be monetized, whether it's sold separately or as part of a bundle, or what pricing structure exists.

• The system is built using Node.js, Express, TypeScript, and integrates with:
  • Telegram Bot API
  • Twilio Voice (with PSTN call initiation)
  • OpenAI GPT-5.6 via Codex
  • ngrok for development
• The author claims that Codex was used to generate or accelerate every line of code during the hackathon.
• Technical challenges included:
  • Speech recognition tuning for Spanish
  • Handling Telegram async flows
  • Ensuring voice consistency in AI agent

Inferred: The use of Codex and GPT-5.6 suggests a low-code or AI-assisted development approach, but no evidence of production-grade infrastructure, scalability, or deployment.

• The project is described as a hackathon submission (OpenAI Build Week 2026).
• It was built in 4 days.
• The author states it is a standalone version of voice tech already used in Dopa, but no evidence of live usage or adoption is provided.
• No customer data, revenue, or user engagement metrics are shared.

Not evidenced: No traction signals such as active users, call volume, conversion rates, or product adoption beyond the author’s own demo.

• The description does not mention any direct competitors.
• It is positioned as a voice automation tool for sales use cases, likely competing with:
  • AI voice agents (e.g., Twilio Voice + AI)
  • Sales automation platforms
  • Conversational AI tools for B2B
• No competitive analysis or differentiation strategy is provided.

Not evidenced: No information on existing competitors, market positioning, or competitive advantages.

• The project is a single-person hackathon effort, with no team or external support.
• It is described as a proof-of-concept and not yet integrated into a live product.
• No evidence of:
  • Revenue
  • Customers
  • Scalability
  • Production-grade infrastructure
• The author claims the system uses GPT-5.6, which is not publicly available, raising questions about technical feasibility or accuracy.
• The use of Codex for development implies a reliance on AI tools that may not be fully reliable or scalable.

Inferred: High risk due to lack of product maturity, traction, and commercial viability beyond the author’s own demonstration.

1. What is the current status of Dopa (the main business)? Is TeleVoice integrated into it or a separate offering?
2. Has there been any real-world testing or customer feedback on the voice agent?
3. What are the technical limitations of using GPT-5.6 in production, especially for voice conversations?
4. How is the system currently being tested or validated?
5. Is there a plan to scale beyond the demo, and what resources are needed?
6. Are there any legal or compliance concerns with AI-generated voice calls?