J'ai laissé une IA négocier avec des inconnus sur Leboncoin

J'ai laissé une IA négocier avec des inconnus sur Leboncoin

by Shubham SHARMA

Applied Sciences & Engineering AI — News & Applications UYComputer scienceUYQArtificial intelligence (AI)
🎙 Shubham SHARMA 👥 313K 📅 November 28, 2024 ⏱ 13 min 👁 161K 📄 documentary 🧭 2026-08-24
Available in: English (current) Français

Keywords

AI voice agentOpenAI Realtime APITwilionegotiationautomation

Summary

In this video, Shubham SHARMA documents his experiment of using an AI voice agent to handle phone calls from potential buyers on the French classifieds platform Leboncoin. He aims to sell his camera equipment without having to personally answer calls. The video walks through the process: crafting a custom prompt for the AI, testing it with a voice agent platform (Vapi), and then moving to OpenAI’s Realtime API to reduce latency. He uses Twilio to obtain a programmable phone number and connects it to the AI via a webhook. After publishing the ad, he receives calls and the AI successfully negotiates, even handling a request for a 25% discount by offering a reduced price. The experiment costs about 20 euros. The creator highlights the impressive naturalness of the conversation and warns viewers about the increasing prevalence of AI voice agents in scam calls. The video concludes with a promotional segment for his training resources.

154 words

Critical Evaluation

Value of the Information & Strength of the Argument

The video provides a valuable, hands-on demonstration of how to build a functional AI voice agent using current APIs. It clearly explains the technical architecture (speech-to-text, LLM, text-to-speech) and the latency issues, contrasting older methods with the newer Realtime API. The argumentation is coherent and based on direct experience, with concrete examples of calls and outcomes. However, the scientific rigor is limited: it is a single anecdotal case, with no controlled experiments or quantitative analysis of negotiation success rates. The creator’s enthusiasm is evident, but the claims about AI’s capabilities are not systematically evaluated.

Scientific Rigor, Source Quality, Title Accuracy

The video cites the tools used (OpenAI Realtime API, Twilio, ElevenLabs) and mentions the book ‘Never Split the Difference’ as a negotiation technique reference. The description provides links to the creator’s own products and training, but no external scientific sources. The title accurately reflects the content, and the video is well-structured with clear chapters. The creator’s technical expertise is credible, but the lack of independent verification or citations to official documentation limits the scientific quality. The video is more of a practical tutorial than a rigorous study.

196 words

Title / Content Match

The title accurately reflects the content: the creator indeed lets an AI negotiate with strangers on Leboncoin.

Quality & Reliability

6/10

The video is a practical demonstration of an AI voice agent for real-world negotiation. It is transparent about the tools used (OpenAI Realtime API, Twilio) and the costs involved, but lacks rigorous scientific methodology or external validation. The creator's expertise is evident in the technical execution, but the content is primarily anecdotal and promotional.

Chapters

Cited Sources

  • Notion Templates — Promotional link for Notion templates, mentioned in the video description.
  • GPT Facile — Promotional link for a resource on using ChatGPT, mentioned in the video description.
  • Formation Make — Promotional link for a training on Make automation, mentioned in the video description.
  • LinkedIn — Creator's LinkedIn profile, mentioned in the video description.
  • Notion Secrets — Promotional link for a Notion training, mentioned in the video description.
  • Notion Facile — Promotional link for free Notion training, mentioned in the video description.
  • Sharma.fr — Creator's website, mentioned in the video description.
  • Automation Facile — Promotional link for a no-code automation training, mentioned in the video description.

Concurring Sources

  • OpenAI Realtime API — Official documentation for the Realtime API, which the video uses to enable low-latency voice conversations.
  • Twilio — Twilio's documentation for programmable voice, which the video uses to set up a phone number and webhook.

Contribution & Novelties

The video offers a practical, accessible demonstration of building a real-time AI voice agent for a common use case (selling items on a classifieds platform). It highlights the technical challenges (latency, voice naturalness) and provides a cost estimate, making the technology tangible for a general audience. The creator’s approach of using the OpenAI Realtime API with Twilio is a concrete example of integrating cutting-edge AI into everyday applications.

Pour aller plus loin :

  • OpenAI Realtime API documentation — Official documentation for the Realtime API, which is central to the video’s technical setup.
  • Twilio Programmable Voice — Twilio’s documentation for programmable voice, explaining how to set up phone numbers and webhooks.
  • Never Split the Difference — Wikipedia article on the negotiation book referenced in the video, providing background on the negotiation techniques used.
  • WebSocket — Wikipedia article explaining the WebSocket protocol, which is used for real-time communication in the Realtime API.

150 words

Radar Profile

The radar profile shows high scores in information quantity and technical level, reflecting the video's detailed walkthrough and practical coding. The quality of information and global reliability are moderate, as the content is anecdotal and lacks external validation. The overall score is solid, indicating a useful but not rigorously scientific resource.

Reliability 5/10

💬 Très positif. Sur les 30 commentaires analysés, la grande majorité exprime de l'enthousiasme et de l'admiration pour la démonstration, avec des remarques humoristiques sur l'ironie de passer plus de temps à automatiser qu'à répondre aux appels, et quelques questions sur le résultat final de la vente.