“Any artificial intelligence smart enough to pass a Turing Test is smart enough to know to fail it.”
Ian McDonald
In a previous Ingenium article we learned about some Project Management Analytics tools and techniques, which are useful for bringing out regularity and support trends in decision-making from project data.
The next step is to derive inferences from data patterns, in order to develop regressive/predictive models through the contribution of AI (Artificial Intelligence). The increasing availability of low-cost machine and deep learning algorithms will shortly lead to the evolution of Project Management software from support systems to decisions for genuine intelligent systems.
What can you do today and what will you soon be able to do with AI to automate an increasing number of project management functions and activities, starting from the project data?
In models of analytics, whether statistical models such as PERT or the Monte Carlo Method, or algebraic models such as the AHP technique, the initial data come from measurements on the project (work performance raw data) or are manually indicated by human experts. In a Monte Carlo simulation you can start from a three-point estimation, which you would probably have done anyway to evaluate project times and costs. And it will always be you who attributes numerical weights to the various decision factors in the AHP technique.
Your evaluations of human experts may derive from experience – so-called “expert judgement” – or from comparison with summary historical data, which can automatically be used for predictive purposes. The basic premise is storage in a historical database of project data in the most detailed and structured way possible (e.g. estimated data, actual data for times, costs and resources used for each activity in previous projects, perhaps grouped by thematic areas or type of project, etc.).
The knowledge contained in historical data can be used to develop a predictive model with Machine Learning (ML) methods, such as neural networks, decision trees, or SVM (Support Vector Machine). Some Project Management domains with greatest ML application are Quality Management (defect prediction), Risk Management and Time & Cost Management. For example, a predictive approach to project times and costs requires the development of a “predictive function or model” through the choice of a specific ML method which, according to historical data, can provide the prediction of times and costs as output, for both the entire project and for specific activities.
In many cases it is the case of a second youth for methods and approaches, once confined in research niches because of their computational complexity, but which have now become feasible thanks to the constant decrease in costs of storage and computing capacity, such as Deep Learning methods, essentially based on neural networks (known from the ’60s) that are much larger and more complex than before.
Analytics + Machine Learning = PMBot?
A bot is an artificial intelligence application which, by greatly simplifying, collects and analyses data, highlighting patterns and imitating – with the aim of improving them – human behaviour. Bots for ordering food or booking a flight online, or chatbots which replace traditional online support systems, are already available.
A bot can be programmed to learn from your best practices. You can use machine learning algorithms to understand how you work, how you analyse problems and how you take decisions, for becoming to all effects an extension of yourself, a genuine virtual assistant.
Imagine you have performed an assessment of project risks. Each risk identified may have an impact on scope, times, costs, change management, etc. A hypothetical PMbot could analyse different scenarios in real or near real time to suggest the best risk mitigation plan based on contextual priorities and best practices. It could simulate the impact of a specific risk on project times and costs, sending the outcome of the analysis to your smartphone while you’re on your way to the planning meeting with your team, helping you to make better and more informed decisions based on an as accurate as possible analysis of data.
Conclusions
Supporting the statistical and algebraic tools of Project Management Analytics, methods and algorithms of machine learning, not new in themselves but now available at increasingly lower costs, your role of “human project lead” will change. Partially freed from your PMbots, you can focus on the non-repetitive and difficult to automate tasks, such as communication with stakeholders and team management. In short, in all those activities where creativity, intuition and your experience as a Project Manager will be irreplaceable for a long time yet. However, to make all this possible, you must prepare yourself in time, collecting and storing your project data with adequate structure and granularity: herein is contained much of the knowledge you need.
Marco Caressa
Recommended reading for learning more
Dinesh Bhagwan HANCHATE, Rajankumar S. BICHKAR (2015). Applied Discrete Mathematics and Heuristic Algorithms Journal, 1, 3 (2015) 21–47. “The machine learning in software project management: A journey. Part I”
Dinesh Bhagwan HANCHATE, Rajankumar S. BICHKAR (2015). Applied Discrete Mathematics and Heuristic Algorithms Journal, 1, 4 (2015) 29–58. “The machine learning in software project management: A journey. Part II”
WanJiang Han1, LiXin Jiang2, TianBo Lu1 and XiaoYan Zhang1(2015). International Journal of Multimedia and Ubiquitous Engineering Vol.10, No.9 (2015), pp.1-8. “Comparison of Machine Learning Algorithms for Software Project Time Prediction”
Project Management Institute (2014). A Guide to the Project Management Body of Knowledge (PMBOK® Guide), 5 th ed. Newton Square, Pennsylvania: Project Management Institute (PMI).
