image AI Single Board Computer - Jetson Nano

Today is a busy day – as is every day in my life as a consulting project and programme manager. I start the day responding to emails and checking the diary, attending video conference calls, checking progress of actions and agreeing new actions.

I then get together with my PMO to review progress updates from our team and suppliers. We update the schedule, and extend our plans into the next phase of work. A team member asks me to review their documents before they send to the client. When that’s done I’m reviewing and refreshing the risk and opportunity register. Every week I prepare for review meetings, compiling and submitting progress reports; much copying and pasting of information throughout. Every month I do the same with all the governance and senior stakeholder reports.

In parallel I’m resolving commercial questions for the business. Will we make the revenue expected on this project? What should the contract look like for the next phase of work? Have we learnt from the past, and are the risks tolerable? I’m allocating roles, coaching, mentoring, and ensuring everyone’s well-being.

And then a substantial risk matures into an issue (such as the Covid-19 pandemic) and everything else pales into insignificance yet cannot be forgotten. All the plates must be kept spinning.

Does this – the life of a project manager – sound familiar to you?

There’s no doubt this life can be frenetic. We recognise that’s part of the excitement and reward of our profession. The challenge and thrill of leading a project through turbulent waters to success.

However, I often think to myself there must be a better way. Can we find more time to breathe, to reflect and guide without having to steer the ship every day? Can we find a way to be less busy, and thereby free ourselves to focus on the big questions and challenges, to maximise our chances of success?

I think we can, based on developments since I wrote an article on AI in Project Management (The Future is Now, AI Is Here to Stay, APM Journal, Summer 2018).

To achieve this goal of ‘less busy, more productive’, I would like to pose a grand challenge to the project delivery community. One which the profession may find uncomfortable, but which ultimately we must embrace, for it is the march of time and technology.

We can and should replace a significant part of what we do, with algorithms: with AI. Such a bold and perhaps contentious idea is now distinctly within the bounds of possibility.

In the last few years the world has moved on and the pace of change is unrelenting. Algorithms and data permeate our society. Unimaginable quantities of data are generated through our myriad daily interactions with online systems. With such pervasive technology, attitudes are changing.

Why should we consider such a radical step?

It is not just about efficiency gains. We owe it to our clients, industry and society to deliver better and more predictable projects. History shows how automation enables economic and societal gains.

However, it’s not a simple picture. How much of our role should we supplant with AI?

There are things a computer does far better than a person. It has perfect recall, is indefatigable, relentless and consistent. It can network and recognise patterns, speech and images. It can construct complex visualisations, can translate between languages, and has unsurpassed numerical capabilities.

But there are things people do far better: we can empathise, we can reason, we can infer, we can abstract and refine, we can induct, recognise and apply patterns across disconnected domains, we can imagine, we can create, we have emotions and form highly complex relationships with others.

So why not combine the best of both?

Let the computer take the strain of repetitive activities within project management. Let it become the autopilot that liberates us to focus on the strategic activities we relish, to be human.

To see how we might achieve this, let’s first characterise what it means to design and operate a project. I think it’s about modelling the world. We scope, design and deliver projects (with varying degrees of success) by building models. Those models manifest in our visions, requirements, plans, estimates, schedules, progress records, reports and – of paramount interest to our clients – outputs and benefits.

Our models reflect the past in the future. Through all our plans, communications and progress reports, we are describing those models and their evolution to our clients. We are telling stories with data.

Some models are descriptive, a statement of what has happened, or how something is. Others are predictive: what we expect to happen. The most powerful automation occurs when we build prescriptive models that not only predict future possibilities but also how to make them happen.

To achieve this we need the best possible models. We must codify our knowledge and expertise accurately, then provide the human oversight that ensures the models function as we expect and benefit from our coaching.

The structured descriptions we need to do this are challenging to build. There are two ways we can approach the problem. We can specify our models using logic, science and hard-won personal knowledge. Alternatively, we can employ process mining and build our models automatically.

Using software tools we can inspect the “digital exhaust plume” of our engagement with online systems to build descriptions of processes and activities as they are actually carried out. These are not theoretical models, but real models.

Having built models of the work, we can employ robotic process automation to accelerate the workflows, allowing one system to automatically drive another system. Data analytics is another burgeoning sector, providing enormous insights into our models of the world that drive new ways of thinking.

A schedule is another type of model: a forecast of what needs to happen to achieve an outcome plus a record of actions to date. The idea that a schedule describes a single path is being replaced with ensembles of paths. Using ensemble modelling techniques we can calculate the probabilities of various outcomes by modelling the likelihood of each potential pathway and its range of efforts and durations.

Similar to chess, a computer now has a clear advantage over a person: it can evaluate countless scenarios rapidly and make optimal recommendations. Ensemble models can be informed by thousands of historical schedules, exploiting hard-won knowledge to improve future outcomes.

This is learning from experience enabled by machine. Our models are not only growing in power and diversity, but they are linking up. The information they contain is no longer sitting in silos. We’re at a tipping point where we can combine them freely. Graph databases play a vital role in this. These are the same engines that power social media sites, and are radically changing how we exchange information. When we join our models up we can generate powerful human-like chains of reasoning.

Process mining can be used to identify where process bottlenecks occur. Then we can pose those scenarios as machine learning problems to find out why these bottlenecks occur. This is applied AI in action – a variety of algorithms being brought to bear on a group of problems to yield human-like insights, but at scale.

Imagine the potential then if we could share project and process data across companies and sectors for the greater good? Data trusts may be the way forward here.

I believe if we combine the best of humans and algorithms, we can reach new levels of project delivery performance. The whole can be greater than the sum of its parts. I anticipate a change in professional skill-sets. Our clients will seek teams who know how to design, build and operate these higher-order systems. Critically, we must understand these systems, their behaviours and the reasons why they reach conclusions – we must never reach the point “computer says no”.

However, I must sound a note of caution. When we replace something, it can only be a part. We must retain control and a full understanding. We are the designers.

My conclusion is that we can indeed begin to meet the challenge. Let us replace a part of our project delivery capability with AI. When we do, our busy days may become quieter and more reflective, as we guide and coach the project systems we design and integrate. Delivery will become more sublime; our clients will enjoy the predictability and confidence we exude as a result.

Perhaps we could then extend this achievement to other professions, and enable greater insights and delivery sophistication throughout society. I believe we can take this opportunity to build a better future.