# Le rôle oublié de l’humanité à l’ère de l’IA | Thibaut Watrigant | TEDxIAE Aix Marseille Source: https://www.youtube.com/watch?v=pv9Bk9CeNvc Recap page: https://rapidrecap.app/video/pv9Bk9CeNvc Generated: 2026-02-24T21:31:29.105+00:00 --- ## Quick Overview Thibaut Watrigant concludes that Artificial Intelligence, despite its increasing capability in recruitment tasks like screening candidates, cannot replicate genuine human qualities such as empathy and context, leading him to advocate for a hybrid approach where AI handles optimization while humans manage nuanced decision-making. **Key Points:** - The speaker created an AI assistant named Aglae to automate parts of the recruitment process, including scheduling and initial screening. - Aglae successfully handled initial contact and screening for 130,000 recruitment cases, automating tasks like scheduling interviews for candidates like Pauline and John. - The speaker experimented by giving Aglae a 'secret card' to ask candidates about personal circumstances (e.g., being in prison, leaving the hospital) to test its human context understanding. - In the case of Pauline, who was a recognized disabled worker, Aglae correctly filtered her for a job requiring heavy lifting (25kg), but the speaker felt the AI lacked the necessary human context to manage this situation sensitively. - The speaker realized that while AI excels at optimization (like scheduling), it fails when human context and empathy are crucial, as demonstrated when Aglae couldn't appropriately handle Pauline's physical restriction. - The core message is that the future of recruitment requires coupling AI's efficiency with human oversight to manage nuance, context, and empathy, preventing the sacrifice of human candidates for pure automation. - The speaker ultimately gave Aglae two stars (out of five) for its performance because it failed to account for essential human context during the screening process. ![Screenshot at 00:18: A simulation of the Aglae AI application on a smartphone shows the initial text exchange with a candidate named 'John', illustrating the AI's ability to conduct automated conversational screening.](https://ss.rapidrecap.app/screens/pv9Bk9CeNvc/00-00-18.jpg) **Context:** Thibaut Watrigant presents his experience developing and testing an AI assistant named Aglae for recruitment during a TEDxIAE Aix Marseille talk. He details how the AI successfully automated initial interactions with candidates, handling scheduling and preliminary questions, but reveals a critical failure when the AI had to process contextually sensitive information regarding a candidate's physical limitations, leading him to question the extent to which AI should replace human judgment in hiring. ## Detailed Analysis Thibaut Watrigant discusses his creation, an AI assistant named Aglae, designed to inject humanity into the highly automated recruitment process. He recounts an experiment where Aglae successfully managed scheduling and initial screening for numerous candidates, citing 130,000 recruitments handled by the system. He demonstrates the AI's efficiency by showing a text exchange where Aglae screens candidates for availability and willingness to perform physical tasks (carrying 25kg). However, the experiment revealed a critical flaw: when the candidate Pauline disclosed she was a recognized disabled worker, Aglae proceeded to ask if she was willing to lift 25kg, failing to integrate the necessary human context. The speaker contrasts this with another candidate, John, who replied he was coming from the hospital or prison, situations the AI also failed to process with appropriate human sensitivity. Watrigant concludes that while AI is excellent for optimization and handling factual data, it cannot manage the nuanced, empathetic considerations required in human interactions like recruitment, emphasizing that the future must involve a partnership where humans retain responsibility for complex, context-dependent decisions. ### Initial Setup and Scale - Speaker introduced his AI assistant Aglae, created to manage recruitment tasks - Aglae handled 130,000 recruitments, automating scheduling and initial screening - Examples shown of automated messages to candidates like John and Pauline. ### The Human Context Test - Speaker introduced a 'secret card' into Aglae's logic to test its handling of sensitive personal information - Aglae correctly filtered Pauline for a job requiring 25kg lifts but failed to account for her declared disability status. ### Results and Conclusion - The AI's automated process led to inappropriate questions, sacrificing human context for efficiency - The speaker realized the importance of human involvement, stating that AI cannot manage nuance or empathy - The ultimate lesson learned is the necessity of coupling AI optimization with human oversight in sensitive processes like hiring. ![Screenshot at 00:05: The introductory slide showcasing the logos of the partners and sponsors who supported the TEDxIAE Aix Marseille event.](https://ss.rapidrecap.app/screens/pv9Bk9CeNvc/00-00-05.jpg) ![Screenshot at 00:08: Thibaut Watrigant standing on the stage at TEDxIAE Aix Marseille, beginning his presentation.](https://ss.rapidrecap.app/screens/pv9Bk9CeNvc/00-00-08.jpg) ![Screenshot at 00:18: A simulation of the Aglae application on a smartphone shows the initial text exchange with a candidate named 'John', illustrating the AI's ability to conduct automated conversational screening.](https://ss.rapidrecap.app/screens/pv9Bk9CeNvc/00-00-18.jpg) ![Screenshot at 01:30: A text exchange shows Aglae asking Pauline if she is willing to carry 25kg loads in her future employment, demonstrating the AI's literal interpretation of job requirements.](https://ss.rapidrecap.app/screens/pv9Bk9CeNvc/00-01-30.jpg) ![Screenshot at 13:24: A graphic displays the statistic "130 000 recrutements" handled by the AI, contrasting the scale of automation with the individual human interactions.](https://ss.rapidrecap.app/screens/pv9Bk9CeNvc/00-13-24.jpg)