If the AI Already Has an Answer: Why We Need to Ask More Detailed Questions

Bonsai BlogShopper ResearchMedia & AnalyticsBonsai Health

Nov. 12

How AI is changing the way we approach pharmacy consultations—and how AI probing helps go beyond initial answers.

By Bettina Mertens-Danowski (Head of Bonsai Health) and Christian Dössel (Head of Bonsai Innovation)

Bettina Mertens-Danowski and Christian Dössel

Anyone who comes into the pharmacy with a question is open to an answer. Anyone who already has an answer may need a different kind of conversation.

This is precisely where one of the challenges of AI-powered preliminary information lies: ChatGPT can help categorize symptoms and prepare questions. However, it can also contribute to turning a mere assumption into a firm expectation. The consultation then begins not with “What’s wrong with you?” but with “Why are you recommending something else to me?” We were interested in this for two reasons:

How does AI-based prior knowledge change self-medication counseling? And what can we learn about this if we don’t stop at the first answer in our research?

bonsAI OTC: When AI Meets AI DOWNLOAD HERE

That’s why we surveyed 80 pharmacy managers and conducted a live pilot of our AI probing tool, bonsAI. AI as a topic, AI as a method—with a technically savvy, time-pressed B2B target audience. We deliberately chose a challenging test case. More information on “bonsAI OTC: When AI Meets AI” is available for download HERE.

Not just more knowledge—but also more certainty

41 of the 80 pharmacy managers surveyed are somewhat or very critical of customers who have researched AI beforehand. 21 take a neutral stance, and 18 are positive. However, it would be too simplistic to conclude from this that “pharmacies are against AI.”

Supporters highlight more targeted questions and the opportunity to build on existing knowledge. Critics, on the other hand, point to specific product requests, a lack of contextual information, and the difficulty of persuading customers who are already convinced to consider a different perspective.

So what matters is not just how well-informed people are, but also how open they remain to receiving advice.

This becomes particularly clear in a pilot dialogue. A pharmacy manager initially justifies her skepticism by saying she does not want to argue against incorrect AI information. When asked, she cites the increased time required. What this means for the consultation only becomes clear upon further inquiry:

"It's less about the patient's problem and more about what the AI says."

This is more than just “AI makes mistakes.” It describes a shift in the conversation: Instead of addressing the individual concern, an existing explanation must first be negotiated.

The pharmacy's strength lies in its ability to address individual cases

This is precisely where the value of professional advice becomes clear: applying general information to the specific case. What symptoms are present, and since when? What medications are being taken? Are there any pre-existing conditions, drug interactions, or warning signs?

The pharmacy managers surveyed identify such gaps in context as significant risks. At the same time, AI’s prior knowledge can serve as a helpful starting point.

For us, this means that preliminary information should neither be accepted without verification nor automatically dismissed. “What have you already researched?” can be just as important as “What are you looking for?”

What This Means for OTC Companies

For self-medication companies, this raises a question that goes beyond traditional product communication: What message comes across when AI selects and summarizes information and links it to a personalized question?

A desire for a particular product may arise even before the pharmacy has clarified the individual context. In a pilot dialogue, a pharmacy manager points out that this desire could even influence the consultation itself: Important questions might “fall by the wayside.”

From our perspective, it is therefore worth exploring this information journey in greater detail. What questions do people ask an AI? What expectations regarding impact and application do they develop as a result? When do they seek expert reassurance—and when do they already consider it unnecessary?

Why We Ask More Detailed Questions Ourselves

Even in research, an initial answer often explains less than it seems. “Misinformation” can refer to incorrect dosages, inappropriate product requests, or a conflict of trust.

This is where bonsAI.live comes in: When input is unclear, the tool asks for clarification and provides more in-depth, context-relevant answers. In the pilot, the number of follow-up questions was intentionally limited; 68 out of 80 cases reached the “Insight” level. This documents the course of the dialogue, but does not automatically reflect the quality of the insights gained.

In the same case, we examined the additional value of the findings: an initial response versus an in-depth dialogue followed by AI-supported summarization. There was no separate control group. The distinction remains important:

Demand uncovers material. Analysis interprets it.

Both require methodological guidance and human judgment. Our pilot study is exploratory, not representative. However, it demonstrates why we are continuing to develop this approach: not to collect as much text as possible, but to better understand what a response means.

It's not the first answer that matters, but the quality of the next question.