Artificial intelligence is fundamentally reshaping how pollsters collect public opinion, with a French emerging company called Naratis leading the charge into what promises to be a faster, cheaper alternative to traditional survey methods. The company, established in 2025 by 28-year-old engineer Pierre Fontaine, utilises conversational AI agents to perform detailed conversations with respondents, replacing the labour-intensive process that has long characterised qualitative research. Rather than requiring respondents to select options, Naratis’s AI engages citizens in genuine dialogue designed to explore not just what they think, but how they think. The technology claims to deliver results ten times faster and at a tenth of the cost of conventional polling, whilst preserving 90 per cent accuracy—a significant breakthrough as the polling industry grapples with declining participation levels and mounting public scepticism.
The Emergence of Conversational Polling
At the heart of Naratis’s innovation lies a seemingly straightforward concept: replacing the transactional character of conventional polling with genuine conversation. When a participant answers the phone, they meet a young, brisk AI voice asking open questions about politics, society and their personal views. Rather than mechanically recording answers, the system engages in real dialogue. Three separate AI agents operate concurrently behind the scenes—one ensuring the respondent remains focused, another probing for further understanding when answers seem superficial, and a third verifying the person is authentic and not a bot exploiting the system. This multi-layered strategy converts polling from a routine box-ticking task into something far more sophisticated and insightful.
The productivity gains are impressive. In the past, qualitative research necessitated lengthy and demanding work: gathering small cohorts of respondents, carrying out one-to-one interviews, documenting spoken exchanges, and then examining answers for patterns and meaning. Naratis compresses the timeframe using what Fontaine describes as “parallelisation”—numerous AI tools performing interviews at the same time rather than interviewers operating sequentially. A study that once took weeks and substantial sums of euros can now be finished in 24 to 48 hours. Responses often arrive by the next day, allowing campaigns, government bodies and groups to address unfolding events and evolving public sentiment with minimal delay, fundamentally changing the tempo of public opinion analysis.
- AI agents conduct concurrent interviews with numerous participants
- Live analysis identifies shallow answers requiring further investigation
- Fraud prevention stops bot activity and dishonest responses from skewing data
- Results provided in just hours as opposed to weeks of conventional research methods
Pace and Effectiveness Transform Survey Methodology
The survey sector confronts an existential crisis. Response rates have collapsed from over 30% in the 1990s to under 5% today, as noted by AI consultant Stéphane Le Brun. This sharp fall has created a vicious cycle: fewer respondents mean increased expenses per completed survey, which in turn renders studies less representative of the wider public. Public trust in polling has eroded accordingly, with many regarding polls as unreliable or intrusive. Against this backdrop, conversational polling powered by AI provides a lifeline, potentially reversing decades of declining engagement by rendering the survey experience itself more engaging and interactive.
Naratis asserts its AI-powered methodology delivers outcomes that are “10 times faster, 10 times more cost-effective and 90% as accurate as human polling.” These statistics, if independently verified, would represent a seismic shift in how organisations understand public opinion. The financial savings alone are transformative: a thorough qualitative investigation that once required tens of thousands of euros and several weeks of labour can now be conducted for a fraction of the price within days. This broader accessibility could allow smaller organisations, grassroots campaigns and community groups to undertake thorough opinion research previously available only to well-resourced organisations.
Parallelisation: The Game-Changer
The innovation enabling these gains is refreshingly simple: parallel processing. Rather than human interviewers performing interviews in sequence—one conversation after another—AI agents function concurrently across dozens or hundreds respondents. This increase in throughput without corresponding cost rises fundamentally alters the economics of polling. Where conventional research methods required considerable time and resources, AI-driven approaches reduce timeframes whilst lowering expenses, enabling companies to gather deep, nuanced insights on demand.
Accuracy Claims and Industry Scepticism
Naratis’s contention that its AI methodology achieves 90% accuracy comparable to human polling has understandably drawn criticism from experienced analysts. The polling industry, founded on decades of technical advancement, remains wary of claims that automated systems can reproduce the subtle discernment of experienced human interviewers. Critics challenge whether conversational AI can truly detect the subtle social cues, hesitations and non-verbal signals that seasoned analysts use to explore more thoroughly respondent motivations. The company has yet to release independent research validating its accuracy claims, leaving independent verification pending.
Beyond concerns about accuracy, sector analysts are concerned about potential biases embedded within AI systems themselves. If the algorithms powering Naratis’s conversational agents are trained on biased data sets or programmed with untested presumptions, those flaws could systematically distort results across thousands of interviews. Additionally, respondents may change their conduct when speaking to machines rather than humans, either becoming more candid or more cautious based on their comfort with technology. These psychological and technical variables are largely unexamined territory, and their effect on polling reliability stays unclear.
- Third-party assessment of accuracy claims is awaiting completion from recognised academic bodies
- Potential algorithmic biases could systematically distort results across large-scale AI polling operations
- AI-human engagement dynamics may influence the way respondents express genuine opinions and beliefs
The Synthetic Data Dilemma
As AI polling scales up, a concerning question emerges: how will regulators and the public distinguish between genuine human responses and synthetic data generated by the very systems running the polls? The efficiency and speed that makes AI polling attractive also opens doors for tampering. If an dishonest actor were to supplement real responses with artificially generated ones, the compiled data could seem statistically sound whilst showing little similarity to actual voter sentiment. The system’s lack of transparency exacerbates the problem—most voters would struggle to understand how algorithms process and verify responses, making it difficult for them to trust the findings driving political conversation.
Naratis asserts its systems incorporate fraud prevention systems, with one AI agent tasked with determining if respondents are real people or automated systems. However, this protective measure itself relies on AI making determinations about AI, producing a circular vulnerability. As dialogue systems become increasingly sophisticated, distinguishing real human exchanges from synthetic responses may be technically unachievable. The opinion research field has traditionally maintained public trust partly because its processes are fundamentally transparent—people provide responses, data are aggregated. AI polling risks compromising that transparency, displacing transparent procedures with opaque algorithms that few can meaningfully audit.
Trust and Regulation Concerns
Regulators throughout Europe are only now come to terms with AI’s involvement in opinion research and political polling. Currently, limited safeguards regulate how AI systems gather, analyse and present polling data. Lacking strong oversight frameworks, the industry faces a crisis of credibility if synthetic data infiltrates published results or if computational biases consistently compromise findings. France’s data protection authorities and the European Union’s AI Act implementation bodies must without delay develop standards ensuring transparency, auditability and accountability in AI-enabled polling work before the technology becomes embedded in political processes.
The Blended Future of Opinion Research
Despite the efficiency improvements AI polling offers, industry specialists suggest that human and machine-driven research will probably coexist rather than one displacing the other entirely. Traditional polling methods have weathered decades of scrutiny and remain integral to political institutions, regulatory frameworks and public understanding. Companies such as Naratis recognise that AI excels at speed and cost-effectiveness, yet human interviewers bring invaluable subtlety—the ability to read subtle emotional cues, adapt questions intuitively and build rapport that promotes candid responses. A balanced approach combining both methodologies could produce deeper understanding whilst maintaining the openness voters increasingly expect from research shaping electoral discourse.
The transition to hybrid models, however, necessitates careful calibration. Pollsters must set out definitive guidelines for the circumstances under which AI data should be given weight alongside traditional responses, and the way conclusions should be shared to ensure the public understands which methods yielded which conclusions. Training a new generation of researchers to collaborate successfully with AI systems poses an additional obstacle, as does establishing professional standards that govern the technology’s application. If approached strategically, this transformation could revitalise opinion research by enhancing efficiency and reach whilst maintaining the human discernment and responsible governance that protect democratic discourse.