Millionways launched ‘Thorsten-4’ to infer what people mean beneath their words, giving sales and hiring teams a way to assess motive from a transcript instead of analyzing only the text itself. The New York startup built the model from a research project in Germany and says it is intended as decision support, not a final verdict.
The training data came from interviews with more than 7,000 people about their lives and goals, then was labeled by psychologists to capture signs of thinking and motive. That grounding can make the model more useful when a team wants a read on how someone may act under pressure or during a high-stakes conversation.
Backed by $6 million in U.S. funding, Millionways gives enterprise buyers a narrower tool than general-purpose AI models. ‘Thorsten-4’ also widens the market for specialized systems that sit within hiring and sales workflows instead of replacing them.
Image Credit: Shutterstock/SuPatMaN
What Makes This Trend Stand Out
- Motivation-aware AI
- Specialized models that infer intent, pressure response, and underlying goals from conversations create room for more nuanced enterprise decision support beyond keyword-based analysis.
- Transcript-based Assessment
- Using recorded interviews and sales calls as structured behavioral data reshapes how organizations evaluate candidates, prospects, and team interactions at scale.
- Psychologist-labeled Models
- Human-expert labeling grounded in behavioral science differentiates narrow AI systems by adding interpretive depth to workplace analytics.
Sectors Adopting This
- Human Resources
- Hiring platforms can incorporate motive and behavioral signal analysis to support richer candidate screening while maintaining human oversight.
- Sales Technology
- Conversation intelligence tools gain new value when they assess buyer intent and emotional drivers rather than only summarizing calls or tracking topics.
- Enterprise AI
- Purpose-built AI systems for defined workflows challenge broad general-purpose models by offering deeper relevance, compliance fit, and operational specificity.
