Conferences & workshops organized by BATVOICE

Part 1 — The market paradox 

Speech Analytics is growing fast, but value concentrates with those who truly personalise. Transforming complex spoken exchanges into actionable insights at scale is a major scientific and operational challenge: identifying intentions, segmenting call stages, extracting decision-relevant indicators while preserving the contextual richness and temporality of human interaction. Our approach is built on an interactive co-design methodology, developed in close collaboration with our clients. AI systems are not designed in a lab and then imposed on the field: they emerge from an iterative process combining business expertise, machine learning engineering, product development and commercial vision. This dynamic produces specific models that are genuinely operational and aligned with sectoral constraints. We also present our hybrid technology approach combining:

• data-driven statistical methods (classification, scoring, segmentation),
• generative models that explain decisions in natural language.

Part 2 — The voice of our clients: Pierre et Vacances Center Parcs 

Panel format. Two clients share their experience on three key points:

• Personalisation: why a generic model was not enough
• Data sovereignty & security : what this means concretely when choosing a partner
• Agility & rapid LLM iteration : how to move from experimentation to ROI in a matter of weeks

Part 3 — What AI should not do: ethics, the EU AI Act and anthropomorphism

Beyond technical performance, we address a central challenge: AI anthropomorphism. As generative models produce natural language summaries, the risk is attributing to them an intention, understanding or judgement they do not possess. This human projection can influence organisational decisions, the perception of results and the responsibility attributed to systems. In this context, we advocate for AI aligned with clear principles: just because a technology is possible does not mean it is desirable. Certain analyses, such as emotion detection, raise ethical and scientific boundaries we choose not to cross.

Part 4 — Looking ahead: analysing human-chatbot interactions 

Finally, we open the discussion towards the evolution of interactions: beyond human-to-human exchanges, how can we analyse and govern human-chatbot interactions without falling into the illusion of symmetry between machine and human interlocutor? This session offers a methodological, technical and ethical perspective on voice analytics in the age of generative AI — placing human responsibility at the heart of the process.