Working to an agreed budget and plan, GREYFLY were commissioned to create a projects prediction proof of concept [PoC]. The PoC was to review existing projects systems and data with the purpose of creating a prediction tool based on pin-pointing learnings from historical project data sources.
The client is a successful IT services business which has a good track record in delivering high profile relatively standardized projects whilst being under acute costs controls.
The assessment was commissioned by the UK Head of PMO who was responsible for supporting project delivery and driving standards and efficiency.
GREYFLY deployed experienced capable consultants and in due course technical specialists who initially investigated existing project systems to identify the maturity of existing data and identify gaps that could enable the AI in PM prediction journey. Findings and recommendations were presented within a report to key stakeholders.
The resultant data comparison was used to inform the ETL [Extract, Transform, Load] stage of ingest. Once data was deemed acceptable priority was given to identify the definition of project success and failure.
These definitions were then used to identify a risk rating template that was then compared to the existing portfolio to provide an individual risk rating.
This data was then used to allocate assurance resources to higher risks projects as well as part of the tool kit when new projects are initiated.
“Based on data points we can now clearly identify which of our projects are highest risk even before they have been mobilised”.
Following on time delivery of the prediction tool proof of concept, the model was accepted, presented to Executives and used to change existing business practices.
Specific insights were found across the projects estate which were then further validated including refining the success and failure blueprint to drive requirements for version 2 of the prediction tool. From a systems perspective the PoC was productionised. However, version 2 is now being developed to further enhance functionality and provide richer data points.