Title : Cost intelligence in the digital oilfield: An AI-enabled framework integrating estimate classification, probabilistic risk analytics, and earned value prediction for capital projects
Abstract:
Capital projects in oil and gas routinely exceed cost and schedule targets by 20 to 80 percent. A 2022 McKinsey analysis of over 500 capital projects found average cost overruns of 79 percent and schedule delays of 52 percent, while Independent Project Analysis (IPA), drawing on a database of over 20,000 projects studied over 35 years, places industrial megaproject failure rates at 65 percent globally. This study demonstrates that these outcomes are predictable and preventable through an integrated Cost Intelligence framework treating estimate governance, probabilistic risk analysis, and earned value management as one technology-enabled discipline.
The framework integrates three components. First, the AACE Cost Estimate Classification System (Class 1–5) operates as a stage-gate governance instrument, enhanced by digital tools mapped to each maturity tier, from big-data benchmarking at early stages to computer-vision-based quantity take-off and cost-loaded digital twins at sanction grade. Second, Cost and Schedule Risk Analysis using Monte Carlo simulation addresses the Narrow Bias problem, where decomposing risks into excessive sub-components produces false statistical precision; this logic extends to disruption and claims quantification through an unsupervised machine-learning Measured Mile technique that removes analyst subjectivity from Loss-of-Productivity calculations. Third, Earned Value Management functions as a prediction system, exploiting the statistical stabilisation of the Cost Performance Index after approximately 20 percent completion to forecast outcomes while corrective action remains affordable.
Applied to a large-scale national energy programme, the integrated framework reversed performance from a Cost Performance Index (CPI) of 0.70 and Schedule Performance Index (SPI) of 0.75 to a CPI of 0.90 and SPI of 0.95 — both unitless performance ratios rather than absolute cost or schedule figures. The recovery was achieved by re-establishing estimate-classification discipline at each stage gate, introducing probabilistic risk analysis where deterministic contingency had previously masked true uncertainty, and redeploying earned value as an early-warning system rather than a retrospective report. Beyond this case, five cost-control maturity heuristics emerged as the practitioner-level indicators that most reliably distinguish mature from immature cost-control environments: scope-led control, separation of the control baseline from approved funding, accrual-based reporting, defensibility of the Estimate at Completion, and disciplined contingency drawdown.
This work contributes a unified, technology-mapped Cost Intelligence framework connecting AI and digital-twin enablement to AACE estimate-maturity tiers, extending probabilistic risk methodology into machine-learning-based disruption and claims analysis, and operationalising earned value management as an early-warning system rather than a retrospective report. All data and methodologies referenced are drawn from public-domain sources, including AACE International Recommended Practices, PMI PMBOK, IPA, and McKinsey Capital Projects research, with no proprietary, confidential, or company-attributable project data disclosed.
