Measurement and optimisation of the performance of probabilistic networks generated by Net AI

Net AI is a novel proprietary technology developed by KADlytics Limited. Net AI can automatically create a network of intra-project dependencies linking digital assets, knowledge and personnel. Net AI's primary market is high-value collaborative engineering projects, such as those in construction, mechanical and software engineering. Engineering projects feature a wealth of dependencies, both explicit and implicit, and understanding the network of dependencies is extremely valuable to Project Managers who need to consider factors such as: Who is responsible for what? Who has knowledge of what? The impact of change? The cost of concessions? It is a lack of understanding of these aspects of high value projects that academic and industrial literature have found to be the common causes for costly failures and overruns.

To this day, the management and subsequent analysis of project dependencies has been manual using surveys and interviews with project members. The time it takes to capture this information results in networks that are immediately out-of-date. Net AI disrupts this practice by generating and updating these networks in real-time and to a higher-level of detail than currently possible. Net AI also makes predictions and recommendations as a project evolves (e.g. determining work allocation when someone leaves, distributing work and knowledge across the company, and evaluating the propagation of change).

Net AI is an emerging state-of-the-art technology and poses a particular challenge in terms of evidencing its performance in terms of: i) coverage of 'real' dependencies and ii) confidence levels in the identification of dependencies as a project evolves. Our initial customers require this understanding and evidence in order to form their internal business case. Similarly, we require this evidence to manage expectations and to quantify the capability of Net AI and hence benefits e.g. 50% of critical dependencies automatically established within the first quarter of a project. This lack of evidence is currently withholding investment decisions by our customers. Since Net AI is a unique AI technology there are no existing off-the-shelf techniques or approaches that can be applied and hence this project endeavours to develop them specifically to measure and optimise the performance of Net AI with respect to i & ii above.

When these techniques are established and effectiveness of Net AI demonstrated cross-industry, it will assist with realisation of potentially huge cost savings for high-value design projects in the UK and world wide.