Imagine you were playing chess for the first time and the opponent just opened the game. It is now your turn, and you are considering what the best strategy would be for your move. The only problem you have is that you don´t know the rules of the game.
Now imagine another scenario. Your are sitting in your car, and type in the address for your target destination in the navigation system. You are told that the location has been found, and now shows a direct line to the destination… but crossing the ocean. Unfortunately your navigation system doesn’t understand what a car is and how that impacts the suggested path towards the goal.
Both scenarios are examples of what happens if a basic set of rules, guidelines or similar input are not present. You would probably get confused, start to guess you way forward, and hope for the best result. Now imagine “Artificial Intelligence” stepping in to help you. Unfortunately, this wouldn’t make any difference as the machine would suffer from the exact same lack of basic information e.g. chess rules and understanding the concept of a car.
So how does the above examples relate to our world of project portfolio management? Well, lets simply imagine that your project managers are all managing projects in the same tool – great start. However, as the organization doesn’t have a Project Model, project work, estimation and tracking is managed completely different. Tasks are given random names, and some have 10.000 tasks while others have 4 tasks. Some create a risk register but only add the name and description of the risk. In this situation, looking only at data, would not be sufficient to identify which project is managed better than the other. It would also be extremely hard to predict if a project is going to finish in time and if risks are under control. Simply exercises like moving unfinished work to start after today’s date, will simply not be accurate enough, as progress might not be reported at a fixed frequency.
In other words, a PPM system in itself is not enough to realize the dream of “Artificial Intelligence” guiding your users and executives in a better direction. One thing is having a PPM database, another is having a common way of working, and finally it all comes down to whether or not users are following the guidelines, or simply ignore them.
With the above introduction, I will try to describe, which options a PMO might have to ensure data and project quality before unleashing the power of “Artificial Intelligence” and “Machine Learning” capabilities. Its much more of a transformational journey than a simple data exercise. I will use the below very simplified matrix to kick start the journey towards “AI”:
I have further simplified things by assuming you are already running Microsoft Project Online or Microsoft Project Server.
No project model or common way of working and therefor nobody following the processes.
In this scenario, your primary task as the PMO/Project Director must be to establish a somehat minimum and common way of working either through developing a “project model” or simply describing a “way of working”. You should also consider what governance that is taking place on a higher level when identifiying, approving and selecting projects for execution. This would often be referred to as the “stage-gate” model.
A project model has been approved by the organization but project managers and executives are not respecting the processes and expected way of working.
Your job is to identify the root cause for this “bad behavior”. Often it has its roots in either an over engineered project model, that is simply to hard to follow, or a badly configured PPM tool, stealing more time from its users, rather than helping them deliver great projects. In other words, motivation for respecting the project processes is not present, and you must figure out a way to remove the “friction” that drives unhappy users. In this case, the use of “BOTs” is a great starting point as these chat-agents, will assist any user, at any point in time, doing things right, leaving them with more time for managing the project progress. Remember that “BOTs” are not necessarily powered by “AI”. In this scenario, a BOT has been given the rules for what to ensure, by who and when. Its more similar to a workflow but with a context based, and more personalized (motivating) approach, to ensure data quality.
A project model is implemented and for the majority of project managers followed and respected.
Congratulations, you have succeeded with something that has required a lot of work, change management and patience. Your organization can now start analyzing the raw data for each process, and start predicting tons of valuable outcomes on almost all levels of your project portfolio management disciplines. Artificial Intelligence and Machine Learning principles will begin to learn and analyze at a pace that would be impossible for humans with a calculator or a spreadsheet. Soon AI will be better than you at defining how the future project model should like like, why some projects deliver better outcomes than others, and predict which projects will run out of money a year from now. Project Managers will experience real time advise e.g. How much time a task should take depending on the actual resource the work has been given to? All powerful input that would surely help any Project Manager deliver a better project, and all powered by Artificial Intelligence, and not an upset PMO worker with a gut feeling.
There is no shortcut from pure planning adhocrazy to smart “Machine Learning” and “AI” predictions. A “BOT” is not necessarily powered by “Artificial Intelligence” but a good starting point. A BOT might as well be a service assistant, that the PMO has trained to understand their requirements and best practices for running a project.
When it comes to Artifical Intelligence and Project Portfolio Management, most are looking for a way to ensure better predictability. But a high degree of accuracy requires a high level data quality, and this requires user discipline. If all Project Managers were truly in need of machine powered predictions in their projects, most would probably be experts in Earned Value and S-Curve visualizations. However, this is rarely the case as most find it too complex and time consuming.
Example of the BOT “PPM Butler” – made by Projectum (www.projectum.com)