Overview

Early-stage ventures live or die by untested assumptions (problem, customer, willingness-to-pay, channels, feasibility). While founders increasingly use generative AI, today’s assistants remain prompt-dependent and tend to produce polished outputs rather than structuring hypothesis-driven learning. This thesis aims to design and evaluate an Assumption Management System that helps founders formulate, discover, prioritize, and evaluate assumptions while preserving founder agency.

Background

Research on AI usage in startups highlights the need for persistent user/context models, coherent intervention logic, and longitudinal support rather than one-off answers. In entrepreneurship, hypothesis-driven approaches emphasize disciplined learning through explicit assumptions and evidence. Yet, in practice, assumptions often stay implicit and scattered across artifacts, making them challenging to track and update. Assumption-based and discovery-driven planning explicitly call for identifying “critical assumptions” and testing them as evidence accumulates. This creates a clear research gap: we lack design knowledge and evaluated prototypes for AI systems that continuously manage an assumption–test–evidence chain in a coaching-aligned way (scaffolding, not substitution).

Potential Research Objectives

This thesis designs and evaluates one integrated artifact that supports assumption-driven analyzation and validation end-to-end. Founders first input and maintain a structured startup context (optionally enriched with venture artifacts such as pitch text or interview notes). The system then analyzes this context to generate and surface implicit assumptions as explicit, falsifiable hypotheses, prioritizes them by uncertainty and expected impact. Finally, it guides founders toward concrete validation strategies by proposing test designs, evidence requirements, and success/failure criteria. The artifact is built and evaluated as a Design Science Research study (prototype instantiation + empirical evaluation against a baseline assistant workflow) to derive design knowledge on how AI systems can scaffold assumption formulation and validation without substituting founder judgment.

Your Profile

• You are interested in the emerging field of generative AI, information systems and entrepreneurship

• You enjoy empirical research (experiments, surveys, interviews)

• You have experience with Python or are willing to learn

• You are highly motivated to work in a self-organized and goal-oriented manner and bring in your own ideas

• Very good English skills, as the thesis will be written in English

We offer an exciting research topic with strong relevance to both academia and practice, close supervision, and the opportunity to develop theoretical, methodological, and practical skills. If you are interested, please send a current transcript of records, a short CV, and a brief motivation (2–3 sentences) to Jonas Liebschner (jonas.liebschner∂kit.edu).