Generative AI in Infrstructure Tendering
A Study on Human-Generative AI Deliberation
R.S. van Duuren (TU Delft - Civil Engineering & Geosciences)
Tong Wang – Mentor (TU Delft - Design & Construction Management)
M. Bosch-Rekveldt – Graduation committee member (TU Delft - Integral Design & Management)
R. K. Soman – Graduation committee member (TU Delft - Integral Design & Management)
More Info
expand_more
Other than for strictly personal use, it is not permitted to download, forward or distribute the text or part of it, without the consent of the author(s) and/or copyright holder(s), unless the work is under an open content license such as Creative Commons.
Abstract
This thesis explores how generative AI (GenAI) can assist in reviewing open infrastructure tenders, focusing on consultants working with contractors. It examines how individuals interact with GenAI, what motivates them to adopt or avoid it, and how these interactions influence the tender process. At the centre is the concept of deliberation, a reflective exchange between humans and GenAI aimed at deepening understanding rather than reaching agreement. The research asks how people experience this process within project-based organisations.
Using a qualitative approach, the study combined a literature review with an experiment involving a realistic tender case. Individuals and GenAI reviewed a tender section, guided by a structured framework. They later discussed their experiences in groups, and industry experts provided further insights.
The Activity Theory was applied, offering a lens to explore how GenAI operates within broader socio-technical systems. Central to this approach is the concept of contradictions, which are understood as tensions within or between components of an activity system that may either act as drivers or barriers for the adoption and use of GenAI. The Human–GenAI Deliberation framework provided structure for meaningful engagement between users and GenAI.
Younger and less experienced participants seemed more open to revising their views, while older users were generally more fixed. Active engagement, including questioning the GenAI, led to more shifts in opinion and was tied to higher trust. However, GenAI’s servile behaviour, rarely doubting the input of the individuals, reduced its impact. Key drivers for use included structured guidance, the objectivity of GenAI, and curiosity reinforced by peers. Barriers involved scepticism, poor prompt formulation, and limited awareness of GenAI’s potential.
The study highlights the importance of building trust and supporting users with clear guidance and training. A major barrier is the lack of knowledge about how to interact effectively with GenAI. Providing users with practical support can build confidence.
Organisations should invest in targeted training, promote informal peer learning, and position GenAI as a collaborative partner. Introducing a GenAI-supported review phase near final tender submission may support deeper reflection. Future research could further validate these findings, explore better prompting strategies, and examine broader applications beyond consultancy.