What would consumers think of a campaign that has not yet occurred? What objections would you have to the new functionality of a mobile phone? How would different audience profiles react to the reformulation of a drink recipe? These are the types of questions to which marketing research has sought answers through surveys, interviews and studies for decades. In the era of artificial intelligence, the marketing industry is evolving towards a new way of finding them: asking consumers who don’t exist.

This is the proposal for synthetic focus groups, a new research modality that uses artificial intelligence models to simulate conversations between virtual profiles of consumers. Instead of recruiting real people to view ads or test products, researchers configure agents with certain demographic, psychological and cultural characteristics, who answer questions, rate stimuli and can interact with each other.

Their function is to emulate the dynamics and information that could be obtained in regular discussion groups, since they generate conversations in which different profiles express opinions, raise debates and react to products, creative concepts, advertising messages or brand strategies. This allows professionals to perform tests, validate hypotheses and explore developments without the need to organize in-person sessions or recruit consumers for each exercise.

That is what is proposed, for example, by Articos, an AI-based research company, based in Dubai and launched in the first quarter of this year. Taking advantage of Apple’s announcement about the launch of its new foldable phone, the iPhone Duo, and the 18 Pro series, the company analyzed how buyers would react.

To do this, he simulated a group of 30 buyers divided into 6 segments and asked them three questions: whether they would make the jump to a Pro model; whether they would be willing to pay $1,999 for the folding model; and what would really motivate them to make a purchase this year. “The results were not what Apple expected with its new range. Most users who, in theory, should have renewed their device did not do so, and no one chose to make the jump to a Pro model”explained the company when sharing the simulation.

Beyond the test results, the video demonstrates the value that the sector can find in this new generation of focus groups: an advance view of how the target audience reacts to a proposal before developing it in depth or launching it on the market. Thus, a marketing department could explore the perception of a price change or identify questions that should be resolved before launching a product; while an agency could present different creative concepts to various synthetic groups, analyze possible interpretations and reformulate the messages before producing the pieces.

In the case of Articos, the profiles are built based on the Big Five model, also known as OCEAN, one of the most adopted psychological models globally and which divides human personality into five large dimensions -Openness, Conscientiousness, Extraversion, Agreeableness, Neuroticism)-, each of which is broken down into 6 facets (30 in total).

In addition, it ensures that the synthetic user profiles are calibrated based on 93 markets so that cultural differences are reflected in the responses; and configure each panel to include diverse personalities, such as defenders, pragmatists, skeptics, detractors and observers, to ensure that there are profiles willing to question the approaches.

Experimentation and savings, among the main advantages

One of the main advantages of this type of focus group is the speed of obtaining information. Organizing traditional discussion groups requires selecting participants, coordinating agendas, preparing stimuli, moderating the sessions and analyzing the results; while synthetics can generate preliminary responses in a matter of minutes.

Cost reduction is also among the benefits. For many brands, especially those with limited budgets, this type of research can be difficult to incorporate into every project. And in this sense, synthetic groups allow us to explore hypotheses that, under other circumstances, might not be investigated. In the case of Articos, the price is between 8 and 20 dollars for the study, significantly lower than the 3,000 – 5,000 dollars for a traditional study, according to the firm’s calculations.

But the most interesting contribution of synthetic focus groups would be found in the ability to experiment continuously. By being able to modify variables and repeat the conversations to observe how the simulated responses change, flexibility and adaptability are obtained that favor research integrated into the work process, and not only concentrated at specific moments.

However, the main obstacle when it comes to the possible adoption of this methodology lies precisely in its nature, that is, the participants in the study are not real people. An artificial intelligence model can imitate some human behaviors, but that does not mean that it provides the same value as the spontaneity, experience, emotionality or unpredictability linked to people.

Furthermore, the dynamics of artificial intelligence could compromise key issues of human communication, such as subtext or non-verbal language. In a conventional focus group, the responses are also determined by the relationships between participants, silences, disagreements or social influence. A conversation between AI agents can represent these interactions, but should not be assumed to reproduce their effects on real people.

On the other hand, synthetic profiles can carry biases from their training data, reinforce stereotypes or generate excessively rational and coherent responses. It is also possible that they reproduced what seems like a likely opinion, or overlooked emerging behaviors that are not yet sufficiently represented in the data.

That is why the research could not be limited to this new modality. To evaluate the effectiveness and value of synthetic focus groups, it is necessary to analyze the extent to which their results coincide with research carried out with real consumers, in what contexts they work best, and what types of questions generate less reliable answers.

Articos itself encourages, in its FAQs section, to combine both modalities and not replace the traditional model with the synthetic one. “While synthetic research is great for testing hypotheses and honing messages at scale, interviews with real people remain the gold standard for uncovering deep emotional nuances and unexpected edge cases later in development.”, they explain.

AI interviewers are also born

But artificial intelligence is not only being applied on the consumer side of the research territory; also on the side of the pollsters. This is the case of Autentic AI, a Spanish start-up with a focus on automating, through artificial intelligence, the complete process of carrying out market studies. As the company itself explains, its platform “Helps teams gather deep insights at scale using conversational AI agents”.

So, instead of manual surveys or interviews, Autentic deploys AI interviewers who hold conversations with users and analyze them to extract themes, emotions, quotes and actionable insights, which, they claim, reduces research time from weeks to hours. What’s more, they ensure that the system is capable of reducing the time-to-insight 6–8 weeks to a few days, or even same-day results could be obtained depending on sample size.

Their proposal, they explain, combines the depth of qualitative interviews with the scalability of quantitative research and can be applied to concept and product tests, pricing research, user journey validation or feedback on messages and positioning.
Unlike Articos, Autentic AI points out in its FAQ section that its platform could replace traditional focus groups and 1:1 interviews, since it is capable of maintaining qualitative depth and contextual richness by eliminating coordination with agencies, long recruitment cycles, manual analysis and high research costs.

At the beginning of October, the company announced that it had obtained 1.5 million euros in a pre-seed financing round in which several venture capital firms participated. He also noted that these funds will be used to boost growth and continue developing the platform.

The figure can be understood as indicative of the industry’s interest in market research solutions based on artificial intelligence, as well as the path they could take. However, its evolution will depend on its true contribution of value to the strategic and creative processes and the way in which it contributes to better research, reserving human contact for those issues in which it remains irreplaceable.