Machine Learning: What Next for Project Management?

Background

Business change projects have a long track record of failure.  Whilst there are clearly identified reasons, the key reason for failure is that they are complex challenges. Implementing business change, without impacting on business as usual (BAU), is difficult. The most complex transformations have been likened to swapping out the engines on a passenger plane while it’s in flight. To extend the analogy, even minor changes to the ‘instruments’ of BAU could have a catastrophic effect. Even if we don’t crash the aircraft (i.e. the business) we can’t be 100% sure that we haven’t changed its behaviour in some imperceptible way, such that long term performance is degraded. AI, i.e. Machine Learning and Predicative Analytics will enable us to minimise that risk and assure delivery.  

Since the environments that businesses and organisations operate are in a constant state of change, i.e. customers and other stakeholders change their needs and expectations, managers and leaders must raise initiatives to make changes big and small. That is, they gather groups of specialists to introduce new technologies, build new business processes and/or build new capabilities. Essentially projects are led by managers who lead people to deliver a specified objective which must be delivered to an agreed budget and on time.

The field and industry of project management emerged in order to manage a collection of people towards an end goal. It’s not uncommon for project managers to manage multiple teams and multiple projects at the same time. These projects (also referred to as programmes) often come together to work towards a wider business objective. Since the advent of the project management specialism, which is infamous for having lots of moving parts, there has been a constant effort to find ways to make the process easier, consistent, and more efficient.

There are two key skills/roles that the Project Manager must master. One is the management and administration of the project which includes the management of financials, reporting, maintaining the plan/schedule and managing project risks. The other role that must be mastered for projects to succeed is that of stakeholder management. The two roles require different skillsets and thinking styles. Management of stakeholders is more of a leadership and people-centric role. Since the attention span of business stakeholders is limited (because they are generally more focused on BAU) project managers must be very effective in this area. Stakeholder management also includes ensuring that the project team, suppliers and other 3rd parties and the business remain focussed on the end goal.   

It is the former ‘administrative’ role where so-called AI has a role to play, allowing project managers and teams to focus on leading and on what people do best.

AI & Project Management

Firstly, we should qualify what we mean by ‘AI’. The first qualification we should make is that we are referring here to a narrow (or weak) AI. That is AI that is specified to handle a singular or limited task. We are not referring to a general (or strong) AI which is at least a generation away from realisation and the capability of which is unknown at this time.

In the “narrow” sense AI has been around for a while. Many of the tasks that the modern smartphone can handle right now would have been described as AI 10 years ago. A number of new and emerging technologies combine to make up Narrow AI and I’ve previously written about them here. It is worth noting that, whilst a number of these technologies are based around the human-computer interface, e.g. natural language processing and image recognition, the heavy lifting of Narrow AI is done by Machine Learning, Deep Learning and Predictive Analytics.  

In the coming years AI, as described above, will increasingly find its way into project management tools and technology. It will handle everything from scheduling of tasks to analysing the patterns of a working team and offering suggestions and tactical and strategic interventions.  

The Art of the Possible: Predictive Analytics for Personalisation

One of the most respected psychological testing tools is the Myer’s Briggs Type Indicator (MBTI). The tool effectively allocated you to one of 16 personality types. Whilst the tool is an excellent guide and is well respected, it remains the case that you will share a personality type with 500 million other souls, which is a pretty broad brush when we’re trying to anticipate how an individual might behave in a particular work scenario. Predictive Analytics, based on the data we capture around how PMs and teams interact with our systems, allows us to generate very precise and accurate ‘nudges’ aimed at very specifically improving individual performance. A system that is capable of analysing a PMs work is likely to be more reliable in predicting actions or potential needs than a human might be. AI holds the key to helping management understand the distinctive nuances that are likely to come with individuals who are working within their own patterns.

Given that we are able to capture much of what project teams produce, i.e. the data of projects, allows us to build out a machine learning model in order to apply predictive analytics against it. This then would allow us to observe the way that a project is moving and make educated predictions about the future of the project. As more projects are predicted and corrected the more accurate the predictions will become. While humans are more likely to get caught focusing on their own problems and may overlook certain shifts in the project because they are not seeing them directly, AI has the ability to watch all of the moving parts and then make valuable predictions based on what it is seeing. AI is capable of monitoring budgets and scheduling, and over time it can learn to identify potential impacts to these processes. 

Over time, it seems likely that AI-enabled project management systems will be able to make the science of human behaviour more concrete in various ways in the same way AI can identify a user profile for a shopper and act accordingly, these systems can help provide customised aid to employees working on a project, as well as project managers. Overall, the potential benefits of bringing AI into the project management space are significant

Where is AI useful?

A 2015 Accenture study indicated that project managers spend over 50% of their time on the administrative tasks of coordination and control. They also conclude that intelligent systems will have a major impact in this area by performing many of these tasks automatically and more rigorously than a human. Essentially making the apparent subjective more objective.   

Use Cases for AI in Project Management

Intelligent PM systems (driven by AI) with its unique ability to monitor patterns will drastically reduce the time that PMs and project teams spend on administration and will ensure that nothing is overlooked. The following is a list of project management task areas where AI will impact:

  • Online templates and forms: capturing the data from online template forms will allow machine learning and predictive analytics across an array of project management functions. Such as
    • Benefits management/realisation: Utilisation of a template-driven business case and benefits tracking tool will allow projects to track benefits delivery and flag actions required where benefits are not being delivered and so suggest appropriate intervention; 
    • RAID (Risks, Assumptions, Issues and Decisions and Dependency) management;
    • Support task/team management by tracking progress against the schedule and highlighting issues;
    • Manage workflows within and between project stages, including gateways and authorisations.  
    • Effectively handle the scheduling of project team members and subject matter experts;
    • Send reminders, and follow-ups and eliminate the need for human input;
    • Cost & time estimation;
    • Budget setting and monitoring;
    • Progress reporting (creation of charts and logs)
    • Effective ‘nudges’ based on an understanding of the project environment and the project manager/ team members working style – aimed at covering areas of weakness.
  • Stakeholder management: monitoring and ‘understanding’ various stakeholder needs and ensuring that the right information is fed to the right people at the right time.

A chatbot might respond to the following stakeholder question like this:

Stakeholder: “how is project ‘Dawn’ doing?

AI Chatbot: “overall it is set to deliver the agreed business benefits on time, but there is some risk that the business case will change because the budget will might need to increase – would you like to know more”

Stakeholder: “yes please”

AI Chatbot: “OK”

AI Chatbot: “The original budget needs to increase by 14.5%, because of the latest change request to allow the additional features that user group identified. The additional budget has not yet been agreed by the project board, but if agreed the payback period agreed in the business case will increase by 4.5 months, do you need to know more?”

Stakeholder: “No thanks – please schedule a meeting for me with the head of the user group and the project manager”

AI Chatbot: “OK – that’s done, it’s tomorrow at 2pm, via Zoom as the Mary (PM) is offsite, is that OK?”

The key to the implementation of the AI-driven intelligent PM systems is integration with popular communication and messaging tools like Outlook and Slack as well as other project management tools like JIRA. Many of these tools are embedded in organisations right now and will be essential in allowing intelligent PM systems to scan existing project collateral and to communicate. In the not-too-distant future, you can expect an AI bot to send you a notification reminding you of some forthcoming project deadline/due date.

Becoming more Human: what will PMs do with the time?

As much of the administrative ‘heavy lifting’ work is done for them, what can PMs do with the time we give them back? Well, they might begin by becoming more human. A PMs effectiveness might previously have been measured by the timeliness of a report, or the level of detail in a risk matrix, or the efficacy of a budget spreadsheet. While the ultimate delivery of the benefits might have been seen as being beyond the PMs control, given the complex nature of organisations, stakeholders and implementation plans. Project focus is generally on getting through the schedule and on delivering a new process or a piece of software. Whilst the exec team wants an outcome, such as a new capability that improves margin, the vast majority of project teams and managers are focussed on delivering the building blocks of the capability, but not the realisation of ultimate benefit. They are generally focussed on things rather than outcomes

AI-driven intelligent project management systems will mean that the skill set of the typical project manager will change. They will need to be less of a manager and more leader, less left-brained (process) and more right-brained (holistic) and less sensing and more intuitive. In the not too distant future, we’ll be able to almost guarantee project success, as the tools at our disposal will flag potential failure very early. PMs will be more people focussed and less process and administrative focussed.  

Quite soon, following the advent of AI-driven project management systems, project managers will have the time and space to be more people and outcome focussed, unencumbered by much of the process PMs will have the time to better support the project team face to face. They will also have more time to think about and solve the higher-order cognitive problems than will assure the delivery of business outcomes.

Postscript: What is Machine Learning good at? A primer

“Machine learning is the science of getting computers to act without being explicitly programmed. In the past decade, machine learning has given us self-driving cars, practical speech recognition, effective web search, and a vastly improved understanding of the human genome” Andrew Ng (Adjunct professor at Stanford University, Co-founder of Google Brain.

Even amongst machine learning practitioners there isn’t an accepted definition of what is machine learning.

Arthur Samuel (1959) describes it as “the field of study that gives computers the ability to learn without being explicitly programmed”, which I think by now we might intuitively all know and accept as a high level concept.

Right back in the 1950’s Samuel’s Checkers (Drafts) board game programme ‘learned’, by playing against itself, what positions led to wins, and losses. Samuels was a poor checkers player and the programme learnt to play better that he could.

This early definition of machine learning has been expanded on by Tom M. Michell (1998). “A computer program is said to learn from experience E’ with respect to some task ‘T’ and some performance measure ‘P’, if its performance on ‘T’, as measured by ‘P’, improves with experience ‘E’.

A great real-world example of the above definition would be an algorithm for classifying emails as spam, or not spam, where

T = Classifying emails as spam or not spam;

E = Watching you label emails as spam or not spam;

P = The number (or fraction) of emails correctly classified as spam/not spam.

There are a number of differing machine learning algorithms, the main ones being:

  • Supervised Learning (where data is labelled, and input and output pairs are known);
  • Unsupervised Learning (looks for previously undetected patterns where data is not labelled;
  • Reinforcement Learning, and
  • Recommender Systems.

Share This:

Post by Pete Wilson

Pete has worked in the technology and business change space for over 30 years. He's worked globally for large public sector and governmental bodies and for large private sector multinationals across numerous industry sectors.

Leave a Reply

Your email address will not be published. Required fields are marked *