I've done a lot of work on machine learning recently and pondered what the future may bring for project and programme managers. Will AI lead to their demise?
According to the Wall Street Journal, organisations believe that the point at which machine learning begins to impact core business functions is rapidly approaching. For those wishing to read further there is a great article in Forbes that helps to explain what an organisation needs to begin its journey into the world of machine learning.
The art of the possible Lets start by imagining a future with AI and then assess how viable it is.
- Core documentation Recent advancements in the legal profession provide a useful insight into what may be coming our way. In the not too distant future routine documentation, strategies, use cases, training plans and security policies could all be auto-populated and adapted from other similar projects. Legacy intranet search capabilities don't always enable us to find the documents we need, when we need them but the latest developments in search will help with reuse and adaptation; leading to significant efficiency gains. Our role will be to add the insights and interpretation as an expert in the field, rather than creating tomes of documentation. We'll be supported by data scientists and analysts who are able to adapt and refine the AI to increase performance and efficiency.
- The plan The days of sitting down and creating a statement of work, plan, budget and risk analysis from a blank sheet of paper will be behind us. As the datasets grow the project manager will have access to a wealth of reference data on which to develop a plan. Iterating from a known set of baselines and adjusting on the basis of key parameters that influence success. Assumption logs will become auto-populated and adjusted for the specifics of a project, identifying which assumptions have the greatest influence on project success.
- AI will also begin to understand the cadence of a project team and forecast the required level of effort. In time, this will also take account of experience (function and sector) and efficiency. This will enable governance bodies to challenge project managers who may seek to empire build rather than demanding the resource profile to deliver the outcome.
- Risk management Machine learning will enable us to capture what risks occurred on which projects, what factors influenced the shape, size and impact of the risk, how likely they are, what impact they had, what mitigation actions were taken and how effective they were. AI will help to shape risk management decision making, whilst also identifying the key parameters that influence their success.
- Decision making AI will help to shape key decisions, informing the probability of a successful outcome by modelling the impact of 'turning left or turning right'. It will take time to capture the volume of data to enable machine learning, but it is certainly achievable. AI will also help to shape responses to specific events.
- Governance Machine learning will be able to advise on the probability of successful delivery within defined parameters, which should influence key investment decisions and prevent ill conceived projects from proceeding. As the reference dataset builds the ability to forecast problems ahead of time will rapidly develop. By automating a rich stream of data, AI will be able to identify areas of focus and help to direct governance interventions. Governance bodies and clients will also be able to interact with 'bots' to interrogate the project and understand the real status.
- Comms and stakeholder management AI could automate a large proportion of stakeholder communications, learning from successful strategies on other projects, adapting templates and ensuring regular comms. Bots will be able to answer questions from clients, stakeholders and aligned projects who may be affected by the project. Project reporting will become automated with the emphasis and shape of the report being adapted depending on the phase, complexity and issues arising. A human in the loop will always be required to ensure that stakeholders requirements are appropriately captured and understood, relationships are nurtured and priorities are captured, but the volume of human effort should diminish.
- Dependency management Bots will also enable projects to talk to each other on a regular basis and flag up areas of misalignment or opportunity. AI can utilise this to prioritise areas of focus, undertake scenario planning and determine courses of action.
- Lessons Learned Machine learning will be used to capture lessons identified/ learned and utilise this knowledge to shape how future projects are delivered. AI will point to specific knowhow, provide checklists and shape recipes for project delivery. Machine learning will be used to assess the potential impact on project performance if lessons are not adopted. It will also enable organisations to connect with others who have delivered similar projects or faced similar challenges.
Project Failure AI will also enable organisations to tackle common reasons of project failure:
1.Enabling effective governance via an unblinking eye, maintaining oversight 24/7.
2.Ensuring that goals and objectives are clearly defined, by comparing to successful and failed projects, whilst also identifying terms which imply inherent ambiguity.
3.Tracking the level of engagement of key stakeholders and tracking potential areas of discontent, friction or organisational politics. Measuring the extent of user engagement.
4.Tracking resource availability, utilisation and competence. Help organisations to address skills gaps and lack of resources, by freeing up capacity for work that is best performed by a human.
5.Identifying how P3M methods influence project success from lessons learned analysis.
6.Benchmarking schedules and assessing probability of success.
7.Assessing requirement stability and the implications of change.
It is unlikely that the role of project manager will ever disappear, but the volume of effort needed to manage a project should rapidly reduce. The extent to which this happens will be driven by a multitude of factors; one of the primary factors will be project complexity. Dave Snowden's work on Cynefin helps to define the boundaries on complexity as follows:
- Simple. Cause-effect relationships are self evident and almost anyone can use them to forecast.
- Complicated. With sufficient time, information and resources we can understand cause-effect relationships and use them to forecast.
- Complex. We can't understand cause-effect relationships in advance but as we see events unfold we can understand how they came about.
In simple and complicated projects AI will help to define the recipes for successful project delivery, balancing the areas of focus dynamically. With a large enough dataset machine learning should be able to model the cause and effect relationships and help to shape decision making. With a dataset of thousands of projects, parameters and environmental influences, AI should be able to inform priorities much better than a human. However, the human will always be required to temper these decisions to accommodate real world constraints, such as a demanding and powerful stakeholder.
In complex projects a recipe based machine learning algorithm is unlikely to succeed. The environment will be too volatile and dynamic, the problem is too difficult to solve. However, AI can help to identify areas of focus and shape decision making by assessing balance of probabilities based on the dataset.
Viability So how viable is this? In my view, it is already with us. The technology is already available and we understand how to apply it. Other professions are applying it but the project delivery profession will be a late adopter.
A significant amount of world has been done by Oxford business school and others on reference class forecasting. The dataset is beginning to evolve.
Tools such as ZiveBox are beginning to emerge which use data to determine task durations and examine the productivity of project team members. Whether this is yet in the realms of AI is for debate. ClickUp is another tool that states that claims to
- Predict the best team member for a task and assign those tasks to them
Automatically tag users in comments based on relevancy contexts
- Predict deadlines that won’t be met
Correct task time estimates
- The foundations are in place, its a matter of the marketplace driving solutions and shaping adoption.
- Predict the best team member for a task and assign those tasks to them
Barriers The key factors that will influence wide scale adoption are:
- Access to data and the quality of the data. There will also be commercial factors which influence how widely data can be shared.
- The complexity of the algorithms to inter-relate a wide range of variables.
- Confidence in the outputs of machine learning, particularly when the datasets are immature. It is easy to dismiss the facts when gut instinct suggests otherwise.
- Competing toolsets that don't have big enough datasets and haven't been sufficiently trained.
I personally view the introduction of machine learning and AI into project and programme management as incredibly exciting. It provides us with an opportunity to improve delivery confidence, improve project outcomes and release the burden of some of the more mundane aspects of our profession. The role of the project manager will always remain; projects will always need someone to manage a team, interface with a wide range of stakeholders whilst balancing their separate requirements and respond to a dynamic environment. But the most effective project managers will be those who are able to exploit the rich seam of data that exists and use this to improve project delivery performance. AI will be the enabler, but it requires a paradigm shift in our profession to harness this power. I welcome your thoughts on how we may be able to achieve this.