Oxford Global Projects was approached by Hong Kong Development Bureau (HKDP) which oversees a large portfolio of infrastructure and building projects. Oxford Global Projects, which was co-founded by Professor Bent Flyvbjerg and Dr Alex Budzier, collaborated with Octant AI to test the ability of AI to predict cost and time over and underruns over the portfolio.
Octant AI provided a reliable detection and early warning systems for the Development Bureau, including the following benefits:
- Reduced holding and opportunity costs for cash through earlier and more accurate prediction of cost and schedule under/overrun performance for all projects within the portfolio.
- Better cost management of projects within the portfolio through earlier and more accurate detection of cost overrun.
- Better schedule management of projects through earlier and more accurate detection of schedule overrun.
- Opportunities for earlier intervention both in treasury and portfolio/project surveillance to enhance HKDP’s portfolio outcomes.
- Reduced time needed to hand-craft statistical and process-related methods for under- and overrun surveillance of any performance indicator.
- Faster, less biased, more succinct and targeted business reporting at a strategic level to reduce noise and latency to make informed business decisions.
- Leverage and liberate historical data for creation of new business and organisational value through powerful and advanced data analytics and artificial intelligence methods.
Octant AI was effective in providing accurate solutions with the predicted results being near-instantaneous.
Scope of Services
Octant AI was appointed by Oxford Global Projects for one week to deliver a cost and time prediction model. This included:
- Data validation and pre-processing
- Cost under and overrun modelling at the project level
- Time under and overrun modelling at the project level
Challenges and Opportunities
- HKDB’s total treasury portfolio is about $USD 40 billion with an annual spend between $USD1 to 2Bn. The challenge has been to accurately provide enough funding each year to cover the dynamics of projects under delivery.
- Oxford Global Projects and Octant AI was commissioned to provide an early warning system to enable accurate prediction of holding costs for their projects and to provide a model for reliable comparison of future performance.
- HKDP’s goal was to improve management of treasury cash flow for its own benefit and for its diverse stakeholders.
Solutions and Innovations
- Under an existing agreement Oxford Global Projects identified the HKDP project as an excellent opportunity for Octant AI to provide a demonstration to augment and complement a conventional statistical analysis.
- Octant AI was validated by and aligned with the excellent statistical analysis carried out by Oxford Global Projects.
- Compared to conventional statistical methods, Octant AI demonstrated a +112% improvement in the overall accuracy of predicting schedule overruns and achieved a +83% improvement in the overall accuracy of predicting cost overruns.
- Octant AI achieved these results by meaningfully using large amounts of complex data sets and producing accurate predictions with a high level of confidence.