Paper Type

Complete

Paper Number

PACIS2026-1535

Description

Public-sector organisations increasingly seek to leverage AI to enhance project portfolio governance. Yet limited research explains how existing governance structures condition AI feasibility at portfolio level. This study asks: How do project portfolio governance structures enable and constrain the effective use of AI for portfolio-level decision-making? Adopting a Walsham-style interpretive case approach, the study integrates strategy analysis, 565 project portfolio artefact examination, and 23 interviews with executive and non-executive stakeholders. Findings demonstrate that AI effectiveness is governance-conditioned rather than technology-determined. Standardised classification and lifecycle structures enable AI-supported analytics, while semantic fragmentation, delivery-centric instrumentation, weak benefits encoding, and magnitude-biased attention constrain portfolio-level optimisation. The study introduces a governance misalignment typology and positions Business Process Management (BPM) as a mechanism for stabilising data ontology and enabling continuous improvement. The findings extend interpretive IS research by reframing AI-enabled portfolio governance as a structural governance redesign challenge rather than a purely technical implementation problem.

Comments

11-Strategy

Share

COinS
 
Jul 5th, 12:00 AM

AI for Project Portfolio Governance in the Public Sector: Opportunities and Limitations

Public-sector organisations increasingly seek to leverage AI to enhance project portfolio governance. Yet limited research explains how existing governance structures condition AI feasibility at portfolio level. This study asks: How do project portfolio governance structures enable and constrain the effective use of AI for portfolio-level decision-making? Adopting a Walsham-style interpretive case approach, the study integrates strategy analysis, 565 project portfolio artefact examination, and 23 interviews with executive and non-executive stakeholders. Findings demonstrate that AI effectiveness is governance-conditioned rather than technology-determined. Standardised classification and lifecycle structures enable AI-supported analytics, while semantic fragmentation, delivery-centric instrumentation, weak benefits encoding, and magnitude-biased attention constrain portfolio-level optimisation. The study introduces a governance misalignment typology and positions Business Process Management (BPM) as a mechanism for stabilising data ontology and enabling continuous improvement. The findings extend interpretive IS research by reframing AI-enabled portfolio governance as a structural governance redesign challenge rather than a purely technical implementation problem.