Faster Cheaper AND Deeper - How Emotion AI Is Transforming Qual Research
Sidi Lamine · June 23, 2025
emotion recognitionfacial expressions
In this session, brand strategist Sidi Lamine explains how integrating emotion recognition technology—via voice tone and facial expressions—enhances qualitative research. The discussion covers the benefits, practical applications, and future potential of AI-moderated interviews for more authentic consumer insights.
0:05 – Nicola Quail:
Welcome everyone. I'm Nicola Quail, co-founder of AI Marketers Guild APAC, joining with colleagues Sushita in Singapore and Dash and Nardi from India. For those new here, the Guild was founded by US marketer David Burkowitz and has grown significantly in North America. We created a regional community to showcase local pioneers in AI tools, marketing best practices, and world-class insights. I'm excited to introduce our guest speaker, Sidi Lamine.
1:09 – Nicola Quail:
Now, let me introduce Sidi Lamine—a globally experienced brand strategist and qualitative researcher with nearly 20 years in leading strategy, research, and innovation across 50 countries. He has supported billion-dollar brands such as Pepsi, Unilever, PNG, Netflix, and Google. Sidi focuses on combining empathy-driven qualitative techniques with cutting-edge AI capabilities to uncover genuine human truths.
2:07 – Sidi Lamine:
Thank you, Nicola. It’s a pleasure to be here. The Marketers Guild has been a fantastic platform for sharing insights. I have been in research, marketing, and brand strategy for nearly 20 years, and I’ve been passionate about AI since studying it back in 2001. When I launched my agency seven years ago, I decided it was time to match my insights with action by integrating advanced tools like emotion analysis.
3:09 – Sidi Lamine:
In qualitative research, we often hear responses like “that sounds great” even if nonverbal cues suggest otherwise. Emotions reveal underlying tensions that competitors might miss, giving us an opportunity to serve our clients better. Missing these emotional cues risks overlooking a critical data point.
4:23 – Sidi Lamine:
If you’ve ever participated in qualitative research, you know that people might verbally express a positive response while their facial cues tell another story. Recognizing these subtle differences is essential to understanding what a participant truly feels.
5:34 – Sidi Lamine:
Nonverbal cues are significant—they reveal when someone struggles to fully express their thoughts. This deeper insight informs us when we need to dig further to uncover any hidden anxieties or hesitations.
6:40 – Sidi Lamine:
Before AI moderation became mainstream, we relied on emotion recognition through voice and facial expression analysis. Voice analysis evaluates tone and pitch to identify key emotions without being dependent on language or accent. Facial expression analysis focuses on micro-expressions that capture subtle, genuine feelings.
7:50 – Sidi Lamine:
AI-moderated research offers incredible benefits—it’s faster, cheaper, and scalable compared to traditional methods. However, using AI alone may sometimes yield flat results if subtle nonverbal cues are lost. Integrating emotion recognition enriches the data by capturing true reactions.
9:00 – Sidi Lamine:
Analyzing voice tone and facial expressions allows us to pinpoint emotional peaks and dips during interviews. This objective data is crucial when presenting findings to senior leadership, offering quantifiable evidence rather than just subjective word clouds.
10:06 – Sidi Lamine:
Voice analysis provides real-time emotional insights, while facial expression analysis detects micro movements that reveal more nuanced emotional states. For example, you might see a spike of happiness during a powerful ad moment and a simultaneous dip in anxiety.
11:05 – Sidi Lamine:
These methods are built on decades of behavioral and psychological research, achieving up to 90% accuracy. They are also privacy safe, with secure, anonymized data processing that scrapes personal information immediately during analysis.
12:09 – Sidi Lamine:
I’ve worked on several case studies—one involving campaign testing and another on creative executions. Small differences in ad presentations can trigger markedly different emotional responses, and recording these subtle shifts in real time gives us invaluable insight.
13:19 – Sidi Lamine:
We often run iterative tests, sometimes with two rounds of participants. With AI capturing live, subconscious reactions, we can clearly see which moments trigger engagement or anxiety. This precise timing is key for improving recall and impacting brand metrics.
14:25 – Sidi Lamine:
Even when participants verbally express neutrality, their voice modulation or facial movements can reveal strong underlying emotions. Such emotional data is essential for refining creative work and making informed strategic decisions.
15:34 – Sidi Lamine:
A common question is whether AI can truly understand emotion. Rather than delve into semantics, I suggest checking out Hume AI’s demo of their voice AI, EVI—which responds empathetically in real time. This demonstration shows that while AI might not “understand” emotion, it effectively detects and reacts to it.
16:45 – Sidi Lamine:
Returning to our case study, we were able to choose a script cut that elicited 25% more joy at critical moments. This wasn’t based solely on what people said—they couldn’t verbally express it—but on verifiable emotional data that directly influenced campaign decisions.
17:49 – Sidi Lamine:
In concept testing, voice emotion recognition is key because it reveals the nuanced differences between similar concepts. Even if participants say both ideas are “fine,” the subtle peaks in their tone guide us in identifying the truly effective approach.
18:52 – Sidi Lamine:
People process stimuli instantly—often within one or two seconds—and those split-second emotional reactions drive their decisions. Capturing these immediate feelings offers more truthful insights than post-interview rationalizations.
19:56 – Sidi Lamine:
Participants might claim neutrality, but the underlying tone of voice can show spikes in joy, anxiety, or anger that matter greatly in decision-making. These instinctive responses are a critical driver of consumer behavior and purchasing decisions.
21:02 – Sidi Lamine:
We apply these insights across ad testing, product development, and exploratory research to uncover unexpected themes. By layering traditional feedback with quantified emotions, we gain a richer, deeper understanding of consumer responses.
22:06 – Sidi Lamine:
Integrating emotion recognition into AI-moderated interviews removes the risk of losing essential emotional insights. This assurance lets us recommend significant innovations with confidence, knowing that the emotional data is robust.
23:11 – Sidi Lamine:
This approach is valuable not just in marketing but also in customer experience and brand health. It allows us to capture true consumer sentiment without the influence of social desirability bias.
24:10 – Sidi Lamine:
In exploratory research, emotion analysis helps identify deep-seated themes and white spaces that traditional methods might miss. It provides a comprehensive picture of consumer behavior that goes beyond what they explicitly state.
25:17 – Sidi Lamine:
The strategy is simple: start slowly, experiment, iterate, integrate, and then scale. Incorporating emotion recognition alongside traditional qualitative methods adds a crucial depth of insight that is often unmatched by competitors.
26:26 – Sidi Lamine:
Once integrated, AI moderation doesn’t sacrifice quality for speed or cost. Instead, it delivers richer, quantified emotional data that directly supports marketing strategies and major client decisions.
27:31 – Sidi Lamine:
Currently, major interview platforms don’t offer built-in emotion recognition. We overcome this by automating workflows using APIs and SDKs, ensuring that you get same-day results without compromising data quality.
28:32 – Moderator:
I have a question—Sidi, you mentioned some tools for running these interviews. Could you share a list of the tools you use for emotion recognition?
29:43 – Sidi Lamine:
For voice emotion recognition, we’ve long worked with a tool from Pho AI; they’re fantastic partners. Additionally, Hume AI offers an API that deconstructs voice into around 42 emotions, though it can be complex. For facial expression analysis, Affectiva leads the field with robust SDKs and extensive datasets. There are several options based on the level of complexity and automation required—feel free to reach out if you need more detail.
30:43 – Moderator:
I have one final question: What opportunities do you see for this technology in enhancing synthetic or virtual personas, especially as we move into the era of AI agents?
31:46 – Sidi Lamine:
The opportunity is immense. With clean, consistent datasets, we can build synthetic personas that accurately emulate human emotions. This would make AI agents much more relatable and effective by enabling them to respond with genuine emotional depth.
32:51 – Dax:
Hi, thanks for the great session. I tested Hume AI about six months ago and explored what could be built on top of it, even experimenting with Speech-to-Text APIs. My main challenge was deploying these solutions quickly for campaign-based activities where marketers need rapid, activation-focused results.
33:58 – Sidi Lamine:
Absolutely, Dax. Hume AI is outstanding for its purpose, but integrating its capabilities into simplified, productized solutions for short campaigns does require additional effort. Its strength lies in long-term, robust applications rather than transient activations.
35:02 – Dax:
I agree—the opportunity for marketeers is to productize and simplify these tools while retaining the depth of emotion recognition. It’s important to strike a balance between advanced technology and practical usability.
36:04 – Sidi Lamine:
Exactly. Integrating qualitative research with AI emotion recognition delivers powerful insights. However, scaling this effectively often requires investment in automation and streamlined workflows.
37:06 – Dax:
That makes sense. In areas where topics are sensitive and truth is hard to capture—like intimate or personal subjects—emotion analysis can uncover what participants might be hesitant to reveal.
38:07 – Sidi Lamine:
This approach is particularly valuable in sensitive areas such as healthcare or intimate products, where participants might guard their true feelings. Emotion recognition helps reveal those deeper insights that are essential for innovation.
39:11 – Dax:
Absolutely. In retail and similar sectors, understanding the underlying emotional responses is proving transformative for how we approach customer engagement.
40:12 – Dax:
As AI agents evolve, incorporating real-time emotional feedback into their training is key to making them more human-like. This integration will improve natural and effective interactions.
41:14 – Dax:
The future will likely see AI agents defined not only by their functionality but also by their emotional intelligence. This will lead to more authentic, engaging, and effective customer interactions.
42:16 – Dax:
If an agent isn’t trained in the nuances of emotion, its interactions can fall short of genuine human connection.
43:19 – Dax:
(Laughs) Protect me from series A and series B discussions! But seriously, if anyone wants to explore our platform further—we have a solution ready, even though it’s not public yet. I have plans to productize this, so please reach out.
44:22 – Sidi Lamine:
Thank you all for joining today. It’s been a privilege to share how AI-moderated qualitative research enhanced by emotion recognition is reshaping our approach to insights. I look forward to further discussions and collaborations—talk to you soon.
0:05 – Nicola Quail:
Welcome everyone. I'm Nicola Quail, co-founder of AI Marketers Guild APAC, joining with colleagues Sushita in Singapore and Dash and Nardi from India. For those new here, the Guild was founded by US marketer David Burkowitz and has grown significantly in North America. We created a regional community to showcase local pioneers in AI tools, marketing best practices, and world-class insights. I'm excited to introduce our guest speaker, Sidi Lamine.
1:09 – Nicola Quail:
Now, let me introduce Sidi Lamine—a globally experienced brand strategist and qualitative researcher with nearly 20 years in leading strategy, research, and innovation across 50 countries. He has supported billion-dollar brands such as Pepsi, Unilever, PNG, Netflix, and Google. Sidi focuses on combining empathy-driven qualitative techniques with cutting-edge AI capabilities to uncover genuine human truths.
2:07 – Sidi Lamine:
Thank you, Nicola. It’s a pleasure to be here. The Marketers Guild has been a fantastic platform for sharing insights. I have been in research, marketing, and brand strategy for nearly 20 years, and I’ve been passionate about AI since studying it back in 2001. When I launched my agency seven years ago, I decided it was time to match my insights with action by integrating advanced tools like emotion analysis.
3:09 – Sidi Lamine:
In qualitative research, we often hear responses like “that sounds great” even if nonverbal cues suggest otherwise. Emotions reveal underlying tensions that competitors might miss, giving us an opportunity to serve our clients better. Missing these emotional cues risks overlooking a critical data point.
4:23 – Sidi Lamine:
If you’ve ever participated in qualitative research, you know that people might verbally express a positive response while their facial cues tell another story. Recognizing these subtle differences is essential to understanding what a participant truly feels.
5:34 – Sidi Lamine:
Nonverbal cues are significant—they reveal when someone struggles to fully express their thoughts. This deeper insight informs us when we need to dig further to uncover any hidden anxieties or hesitations.
6:40 – Sidi Lamine:
Before AI moderation became mainstream, we relied on emotion recognition through voice and facial expression analysis. Voice analysis evaluates tone and pitch to identify key emotions without being dependent on language or accent. Facial expression analysis focuses on micro-expressions that capture subtle, genuine feelings.
7:50 – Sidi Lamine:
AI-moderated research offers incredible benefits—it’s faster, cheaper, and scalable compared to traditional methods. However, using AI alone may sometimes yield flat results if subtle nonverbal cues are lost. Integrating emotion recognition enriches the data by capturing true reactions.
9:00 – Sidi Lamine:
Analyzing voice tone and facial expressions allows us to pinpoint emotional peaks and dips during interviews. This objective data is crucial when presenting findings to senior leadership, offering quantifiable evidence rather than just subjective word clouds.
10:06 – Sidi Lamine:
Voice analysis provides real-time emotional insights, while facial expression analysis detects micro movements that reveal more nuanced emotional states. For example, you might see a spike of happiness during a powerful ad moment and a simultaneous dip in anxiety.
11:05 – Sidi Lamine:
These methods are built on decades of behavioral and psychological research, achieving up to 90% accuracy. They are also privacy safe, with secure, anonymized data processing that scrapes personal information immediately during analysis.
12:09 – Sidi Lamine:
I’ve worked on several case studies—one involving campaign testing and another on creative executions. Small differences in ad presentations can trigger markedly different emotional responses, and recording these subtle shifts in real time gives us invaluable insight.
13:19 – Sidi Lamine:
We often run iterative tests, sometimes with two rounds of participants. With AI capturing live, subconscious reactions, we can clearly see which moments trigger engagement or anxiety. This precise timing is key for improving recall and impacting brand metrics.
14:25 – Sidi Lamine:
Even when participants verbally express neutrality, their voice modulation or facial movements can reveal strong underlying emotions. Such emotional data is essential for refining creative work and making informed strategic decisions.
15:34 – Sidi Lamine:
A common question is whether AI can truly understand emotion. Rather than delve into semantics, I suggest checking out Hume AI’s demo of their voice AI, EVI—which responds empathetically in real time. This demonstration shows that while AI might not “understand” emotion, it effectively detects and reacts to it.
16:45 – Sidi Lamine:
Returning to our case study, we were able to choose a script cut that elicited 25% more joy at critical moments. This wasn’t based solely on what people said—they couldn’t verbally express it—but on verifiable emotional data that directly influenced campaign decisions.
17:49 – Sidi Lamine:
In concept testing, voice emotion recognition is key because it reveals the nuanced differences between similar concepts. Even if participants say both ideas are “fine,” the subtle peaks in their tone guide us in identifying the truly effective approach.
18:52 – Sidi Lamine:
People process stimuli instantly—often within one or two seconds—and those split-second emotional reactions drive their decisions. Capturing these immediate feelings offers more truthful insights than post-interview rationalizations.
19:56 – Sidi Lamine:
Participants might claim neutrality, but the underlying tone of voice can show spikes in joy, anxiety, or anger that matter greatly in decision-making. These instinctive responses are a critical driver of consumer behavior and purchasing decisions.
21:02 – Sidi Lamine:
We apply these insights across ad testing, product development, and exploratory research to uncover unexpected themes. By layering traditional feedback with quantified emotions, we gain a richer, deeper understanding of consumer responses.
22:06 – Sidi Lamine:
Integrating emotion recognition into AI-moderated interviews removes the risk of losing essential emotional insights. This assurance lets us recommend significant innovations with confidence, knowing that the emotional data is robust.
23:11 – Sidi Lamine:
This approach is valuable not just in marketing but also in customer experience and brand health. It allows us to capture true consumer sentiment without the influence of social desirability bias.
24:10 – Sidi Lamine:
In exploratory research, emotion analysis helps identify deep-seated themes and white spaces that traditional methods might miss. It provides a comprehensive picture of consumer behavior that goes beyond what they explicitly state.
25:17 – Sidi Lamine:
The strategy is simple: start slowly, experiment, iterate, integrate, and then scale. Incorporating emotion recognition alongside traditional qualitative methods adds a crucial depth of insight that is often unmatched by competitors.
26:26 – Sidi Lamine:
Once integrated, AI moderation doesn’t sacrifice quality for speed or cost. Instead, it delivers richer, quantified emotional data that directly supports marketing strategies and major client decisions.
27:31 – Sidi Lamine:
Currently, major interview platforms don’t offer built-in emotion recognition. We overcome this by automating workflows using APIs and SDKs, ensuring that you get same-day results without compromising data quality.
28:32 – Moderator:
I have a question—Sidi, you mentioned some tools for running these interviews. Could you share a list of the tools you use for emotion recognition?
29:43 – Sidi Lamine:
For voice emotion recognition, we’ve long worked with a tool from Pho AI; they’re fantastic partners. Additionally, Hume AI offers an API that deconstructs voice into around 42 emotions, though it can be complex. For facial expression analysis, Affectiva leads the field with robust SDKs and extensive datasets. There are several options based on the level of complexity and automation required—feel free to reach out if you need more detail.
30:43 – Moderator:
I have one final question: What opportunities do you see for this technology in enhancing synthetic or virtual personas, especially as we move into the era of AI agents?
31:46 – Sidi Lamine:
The opportunity is immense. With clean, consistent datasets, we can build synthetic personas that accurately emulate human emotions. This would make AI agents much more relatable and effective by enabling them to respond with genuine emotional depth.
32:51 – Dax:
Hi, thanks for the great session. I tested Hume AI about six months ago and explored what could be built on top of it, even experimenting with Speech-to-Text APIs. My main challenge was deploying these solutions quickly for campaign-based activities where marketers need rapid, activation-focused results.
33:58 – Sidi Lamine:
Absolutely, Dax. Hume AI is outstanding for its purpose, but integrating its capabilities into simplified, productized solutions for short campaigns does require additional effort. Its strength lies in long-term, robust applications rather than transient activations.
35:02 – Dax:
I agree—the opportunity for marketeers is to productize and simplify these tools while retaining the depth of emotion recognition. It’s important to strike a balance between advanced technology and practical usability.
36:04 – Sidi Lamine:
Exactly. Integrating qualitative research with AI emotion recognition delivers powerful insights. However, scaling this effectively often requires investment in automation and streamlined workflows.
37:06 – Dax:
That makes sense. In areas where topics are sensitive and truth is hard to capture—like intimate or personal subjects—emotion analysis can uncover what participants might be hesitant to reveal.
38:07 – Sidi Lamine:
This approach is particularly valuable in sensitive areas such as healthcare or intimate products, where participants might guard their true feelings. Emotion recognition helps reveal those deeper insights that are essential for innovation.
39:11 – Dax:
Absolutely. In retail and similar sectors, understanding the underlying emotional responses is proving transformative for how we approach customer engagement.
40:12 – Dax:
As AI agents evolve, incorporating real-time emotional feedback into their training is key to making them more human-like. This integration will improve natural and effective interactions.
41:14 – Dax:
The future will likely see AI agents defined not only by their functionality but also by their emotional intelligence. This will lead to more authentic, engaging, and effective customer interactions.
42:16 – Dax:
If an agent isn’t trained in the nuances of emotion, its interactions can fall short of genuine human connection.
43:19 – Dax:
(Laughs) Protect me from series A and series B discussions! But seriously, if anyone wants to explore our platform further—we have a solution ready, even though it’s not public yet. I have plans to productize this, so please reach out.
44:22 – Sidi Lamine:
Thank you all for joining today. It’s been a privilege to share how AI-moderated qualitative research enhanced by emotion recognition is reshaping our approach to insights. I look forward to further discussions and collaborations—talk to you soon.