AI Probing: From promise to practice 

In August, we introduced you to AI probing and its potential to transform open-ended survey questions from shallow one-liners into rich, layered responses. We outlined how AI-driven follow-up questions could unlock depth at scale, particularly when aligned with frameworks like Jobs to Be Done. 

Three months on, the use of Natural Language Processing (NLP) tools that generate context-aware follow-up questions within surveys is becoming common practice. And we’ve discovered something even more powerful: AI doesn’t just help us ask better questions during data collection; it fundamentally changes how we analyse what we hear. 

What we’ve learned since August 

The collection side has exceeded expectations. Studies using AI-driven iterative probing platforms are seeing completion rates above 90% across all prompts. Respondents aren’t just tolerating the follow-up questions; they’re engaging deeply, providing an average of three to four detailed responses per person. 

More significantly, we’re capturing genuinely multi-dimensional reasoning. The average respondent articulates two or more distinct motivations for their choices. Roughly 10% emerge as particularly expressive, weaving together five or more attributes that combine functional benefits, emotional drivers, and social validation in ways that reveal the full complexity of decision-making. 

Here’s what that looks like in practice: “This brand gives me complete peace of mind. My mechanic has recommended it for years, and I can genuinely feel the difference in how the car performs. It costs more upfront, but it’s worth it because it protects the engine and saves money on repairs in the long run.” 

That’s not a tick-box response. It’s trust earned through expert validation, performance experienced firsthand, and value calculated over time, captured at quantitative scale. We’re getting hundreds of responses like this per study, not just a handful from focus groups. 

The analysis breakthrough 

The real evolution though has been on the analysis side. A typical study can now generate over 2,500 individual text entries. We’ve moved beyond semi-automated coding to using both general AI platforms and research-specific AI tools for sophisticated interrogation of this verbatim data. 

With well-designed prompting, we can ask AI to segment respondents by narrative complexity, identify which competitors own which emotional territories, or flag unexpected patterns. We can request: “Show me every instance where trust and performance appear together” or “What are the common narrative structures among high-engagement respondents?” The AI retrieves and organises the evidence; our team interprets what it means strategically. 

We analyse at both macro level (category-wide themes) and micro level (individual narratives that reflect key segments). This dual perspective reveals not just what themes exist, but how different customer segments articulate them. 

Critically, we validate every AI-generated insight against actual source text. This isn’t about letting algorithms imagine patterns; it’s about using AI to surface and organise evidence more quickly, while our researchers handle strategic interpretation. Every finding traces back to what respondents genuinely said.  

The ROI equation has changed 

Compare traditional quantitative research (“62% agree this brand is trustworthy”) with what we’re now delivering: “Trust is earned through three distinct pathways: expert validation from mechanics or enthusiasts, performance differences in everyday use, and long-term protection that prevents costly failures.” 

That specificity changes everything. Segmentation becomes richer: you can group customers by how they think and decide, not just demographics. Brand positioning gains precision: you identify the multi-attribute stories that genuinely drive preference. And you gain authentic customer language for messaging development and creative briefs. 

The methodology works across brand health tracking, category entry studies, positioning development, customer journey mapping, and NPS driver analysis that goes beyond the score to understand the story. That’s qualitative depth at quantitative pricing and speed. And it’s no longer theoretical; it’s how we’re designing research today. 

The question we posed in August was whether AI probing could transform insight at scale. Three months of practice have given us the answer: it already has. 

Interested in using AI probing in your next project?

Want to explore how AI probing could work in your next market research study? Contact us to learn more about how we can integrate it into your research design.