Artificial Intelligence and Organisational Viability in Early Stage Venture Capital

Abstract

Early-stage venture capital (ESVC) decision-making operates under conditions of persistent uncertainty, partial visibility of work, and long feedback delays between investment decisions and outcomes. While recent attention has focused on the potential of artificial intelligence (AI) to improve investment decision-making, less attention has been given to how AI interacts with the organisational structures through which ESVC decisions are produced and regulated. This thesis addresses that gap by examining ESVC decision-making as a question of organisational viability rather than one of optimisation or individual judgement quality. Its aim is to diagnose how ESVC decision systems are functionally structured, how persistent structural tensions within those systems are managed across historical and contemporary practice, and what these tensions imply when AI alters the informational environment in which organisational regulation occurs.

Drawing on organisational cybernetics, and in particular Beer's Viable System Model (VSM), this research adopts a diagnostic, abductive approach to analyse how ESVC decision systems manage complexity over time. The empirical study focuses on a single, bounded regional innovation ecosystem characterised by proximity to research institutions, deep-technology ventures, and dense relational networks among investors, intermediaries, and public agencies. Three analytical lenses structure the empirical work. The first examines a historical baseline through analysis of an anonymised ESVC archival corpus spanning eighteen years. The second examines contemporary practice through semi-structured interviews with ESVC practitioners, ecosystem orchestrators, and technology specialists. The third synthesises these lenses to identify persistent structural tensions and interpret their implications under altered informational conditions, specifically, the introduction of AI.

The archival diagnosis shows that all core VSM subsystems were present but unevenly enacted, giving rise to structural tensions across four dimensions: autonomy and representability in operational work, formality and informality in coordination, authority and access in oversight, and discretion and consistency in normative governance. Practitioner accounts confirm that these tensions persist in contemporary practice despite changes in tools and organisational contexts, and that they are managed through informal judgement, trust-based calibration, and situational coordination rather than formal regulatory capacity.

AI does not resolve these tensions but reconfigures the informational conditions under which they are managed, through four system-level effects: attenuation of local effort, amplification of interpretive misalignment, redistribution of regulatory pressure across roles and recursive levels and obscuring of tacit conditions under which decisions are produced and legitimised. This diagnostic reframes the regulatory function of features (trust, heuristics and informal judgement) that are commonly framed in the decision-making literature as sources of decision error.

This thesis makes several contributions: it reframes ESVC decision-making as a question of organisational viability, positioning judgement, trust, heuristics, and bias as functional responses to structural conditions; it provides the first VSM diagnostic of ESVC decision-making at fund level, combining longitudinal archival and contemporary practitioner data; and it conceptualises AI as a redistributor of informational strain rather than a decision-maker or corrective, contributing a diagnostic lens for analysing how AI alters the conditions under which coordination, oversight and organisational learning occur in complex decision systems characterised by uncertainty and delayed feedback.