17 November 2026
AI closed the last gap in learning about customers
This is a continuation of Knowledge about talking to customers was always free enough (opens in a new tab).
Free knowledge left one gap: someone who answers your question about your situation, patiently, whenever you ask. Education research has had a name for that gap for forty years.
The two sigma problem
In 1984 Benjamin Bloom reported that students taught one-to-one by a tutor did far better than students in a normal classroom and he called the question of how to give everyone that advantage the 2 sigma problem [1]. The problem was cost. A personal tutor for every learner did not scale, so most of us made do with books and rooms.
A tutor for everyone
An AI assistant answers at any hour, in your words and about your case. A randomized controlled trial in a Harvard physics course with 194 students compared a custom AI tutor with an active learning class: the median learning gains of the AI-tutored group were more than double those of the group in class [2]. The tutor was designed with prompt engineering and teaching practice and was not a plain chatbot.
The limits
That is one course at one university, with a tutor built for the purpose, and the authors themselves do not claim that AI tutoring will always beat the classroom [2]. A general assistant can be wrong and it only knows the method, never your customers. Treat the evidence as early and promising.
What this looks like for customer interviews
The flyer on this site contains a prompt that turns any AI assistant into a tutor for customer conversations: it builds a syllabus, explains one concept at a time, asks you to apply it to a purchase you made yourself and pushes back when your answers are shallow [3]. About an hour and a half with it gets you through the concepts that used to take a shelf of books.
What it changes for a team
"We did not know how" stops being a reason. A team can learn the method in an afternoon and the question becomes what is still hard. Two things are and no tutor removes them: talking to real people and practising the skill. The next two posts take them one at a time.
You can practise this against an AI customer in the app (opens in a new tab), where nobody real is at risk, or grade a recording (opens in a new tab) of an interview you already did. The first practice round is free and so is grading a recording.
References
- Wikipedia. Bloom's 2 sigma problem. https://en.wikipedia.org/wiki/Bloom%27s_2_sigma_problem (opens in a new tab)
- Kestin, G., Miller, K., Klales, A., Milbourne, T. and Ponti, G. (2025). AI tutoring outperforms in-class active learning: an RCT introducing a novel research-based design in an authentic educational setting. Scientific Reports, 15, 17458. https://doi.org/10.1038/s41598-025-97652-6 (opens in a new tab)
- Customer Conversations. The 28-page flyer with the learning prompt. https://www.customer-conversations.com/en/flyer (opens in a new tab)