Generative AI can cut project time and budget slippage and boost quality—especially in knowledge-intensive projects; gains are modelled at roughly 10–20% but depend heavily on industry context, data quality and staff training.

Main Finding

Generative AI tools (e.g., ChatGPT, Copilot, Claude) materially improve project management efficiency and financial performance across project life‑cycle stages. The paper estimates typical gains of ~10–15% reduction in execution time, 10–20% reduction in budget deviations, and improved ROI—effects strongest in knowledge‑intensive projects (IT, consulting) and moderate in construction. AI complements (does not replace) classical methodologies (Agile, Waterfall) by automating routine tasks, improving forecasting/risk detection, and strengthening financial discipline.

Key Points

• Quantified impacts reported:
  • Execution time reduced by ~10–15%.
  • Budget deviations decreased by ~10–20%.
  • Many adopters report positive ROI within a year (cited survey evidence).
• Sector heterogeneity:
  • Largest financial gains in IT and consulting (data‑driven decision making, automation of knowledge work).
  • Moderate gains in construction (improved planning and risk mitigation).
  • Benefits also noted in education and research (faster drafting, documentation, reporting).
• AI use across life cycle:
  • Initiation: drafts project charters, clarifies requirements, improves cost‑benefit appraisal.
  • Planning: automated schedules, WBS, risk identification, resource optimization.
  • Execution: automates routine tasks (status updates, drafting, code suggestions), reduces errors/rework.
  • Monitoring: real‑time progress tracking, early deviation warnings, automated managerial summaries.
  • Closing: automated lessons learned, final reports, better post‑project cost tracking.
• Role relative to methodologies: additive — strengthens Agile/Waterfall by improving planning, forecasting, and execution without displacing core frameworks.
• Risks and constraints: data privacy, ethical/legal issues, integration barriers, need for reliable data and staff training.

Data & Methods

• Multi‑method approach:
  • Literature review of surveys, case studies, and prior research on AI in project management and finance.
  • Comparative scenario modeling for five representative project types: IT, construction, consulting, education, research. Each had two scenarios: traditional vs AI‑supported.
  • Financial indicators analyzed: ROI, cost variance, schedule performance, frequency of budget overruns.
  • Data sources: industry surveys (APM, Project Management Statistics), published case studies, company reports, and expert opinion—no single comprehensive dataset.
  • Fuzzy logic model to handle uncertainty: defined improvements ΔT (time), ΔC (cost), ΔQ (quality) as fractional changes (Δ = (without − with)/without), mapped to linguistic categories (Low/Medium/High) and evaluated via if–then fuzzy rules to produce an overall impact rating (defuzzifiable to an index).
• Assumptions and limits:
  • Impact treated as additive to existing methodologies.
  • Many quantitative estimates derived from synthesized survey/case evidence and expert judgments rather than controlled experimental data.
  • Thresholds for fuzzification (e.g., ~5% = Low, ~15% = Medium, >20% = High) are adjustable and based on expert consensus.

Implications for AI Economics

• Microeconomic/project‑level:
  • AI adoption can raise project productivity and lower realized project costs, implying higher ex‑post ROI for firms that implement AI correctly.
  • Sector heterogeneity matters: models of AI’s economic impact should differentiate knowledge‑intensive vs capital‑intensive projects (larger elasticities in the former).
  • Labor effects are likely complementary/augmentative for skilled project staff (automation of routine tasks), reducing billable hours for routine work and shifting labor to higher‑value tasks—important for firm wage and billing models.
• Firm strategy and investment:
  • Firms should factor AI‑enabled cost savings and risk reduction into project selection, capital budgeting, and portfolio optimization (reduced probability of overruns changes expected project NPV distributions).
  • Investment needs: upfront spending on tools, staff training, data governance—costs that must be weighed against projected operational savings and faster payback.
• Public policy and procurement:
  • Public project appraisal and procurement frameworks should consider AI’s potential to reduce cost variance and schedule risk; however, procurement must also ensure data/privacy compliance and vendor transparency.
• Measurement and research agenda:
  • Need for standardized, longitudinal datasets (project‑level time, cost, quality metrics before/after AI adoption) to credibly estimate causal effects and heterogeneity.
  • Randomized pilots or difference‑in‑differences studies in organizations would strengthen evidence on effect sizes and persistence.
  • Macro modeling: aggregate productivity models should incorporate heterogeneous firm‑level uptake and sectoral differences; sensitivity analyses to data quality and integration costs are necessary.
• Cautions for economists and managers:
  • Reported gains rely substantially on surveys and expert syntheses; beware of selection/response bias—early adopters may be atypical.
  • Non‑technical constraints (data governance, ethics, change management) can materially reduce realized gains; policy and organizational investments are required to capture the projected financial benefits.

Summary takeaway: Generative AI is a strategic driver of project‑level financial efficiency, especially for knowledge work, but robust measurement, attention to sectoral differences, governance, and investment in complementary capabilities are essential to realize and credibly quantify those economic gains.