Ask Onix
AI transforms political polling with conversational interviews
A Paris-based technology firm, Naratis, is redefining political opinion research by replacing human interviewers with artificial intelligence. Founded in 2025 by 28-year-old engineer Pierre Fontaine, the company conducts in-depth qualitative polls through AI agents that engage respondents in real-time conversations.
How Naratis works
When a participant calls Naratis, an AI agent initiates a dialogue, such as asking what image or emotion comes to mind when hearing the word "politician." While the respondent answers, three additional AI systems analyze the interaction: one ensures the response addresses the question, another evaluates depth, and a third verifies the participant's authenticity to prevent fraud.
Fontaine emphasizes the method's focus on qualitative insights rather than quantitative data. "We don't ask people to tick boxes-they have a conversation with an AI," he says. "This allows us to explore not just what people think, but how they think."
Speed and cost advantages
Naratis claims its approach is "10 times faster, 10 times cheaper, and 90% as accurate as human polling." Traditional qualitative studies, which involve small groups or one-on-one interviews, can take weeks and cost tens of thousands of euros. Naratis completes similar projects in one to two days, with data collection often finished within 24 hours.
The efficiency stems from "parallelisation"-AI agents conduct multiple interviews simultaneously, unlike human interviewers who work sequentially. This rapid turnaround enables clients to respond to events in near real-time.
Industry challenges and AI's role
The polling industry faces declining response rates, which have dropped from over 30% in the 1990s to below 5% today, according to AI consultant Stéphane Le Brun. Lower participation increases costs and reduces representativeness, fueling public skepticism.
Fontaine argues that AI can mitigate these issues, particularly in qualitative research. Unlike quantitative polls-often criticized for high-profile misses like Brexit or Donald Trump's 2016 victory-qualitative studies aim to uncover nuanced opinions, such as reactions to campaign slogans, rather than predict election outcomes.
Broader adoption and ethical concerns
Established firms like Ipsos are also integrating AI into polling. Techniques include analyzing social media, using "digital twins" (virtual models of real individuals), and generating synthetic data-entirely new profiles based on real-world patterns. These tools help study hard-to-reach groups but raise ethical questions.
Bruno Jeanbart, CEO of OpinionWay, warns against over-reliance on AI-generated data. "We would never publish an opinion poll based on synthetic respondents," he says, citing trust concerns. Ipsos similarly avoids using AI-generated participants in political surveys.
Critics highlight risks like AI "hallucinations"-plausible but incorrect responses-and the potential for synthetic data to distort public opinion measurement. Le Brun notes that while full automation is the goal, "it would be unsafe and socially unacceptable to remove humans entirely."
Future of polling: hybrid models
The most plausible path forward combines AI efficiency with human oversight. AI enables large-scale conversational surveys, faster insights, and reduced bias-participants may be more candid with machines, particularly on sensitive topics like far-right support in France.
However, the boundary between augmenting human data and simulating it remains critical. Naratis and similar firms are betting that the real innovation lies in transforming surveys into dynamic conversations, not replacing respondents.
"Whether this shift restores trust in polling or further erodes it will depend less on the technology itself than on how it is used, explained, and regulated,"
Industry analysts
Economic pressures will likely accelerate automation, but the industry's credibility hinges on transparency and ethical safeguards.
