What does trust mean in the context of artificial intelligence in a project management setting?

For readers who aren’t project or programme management specialists, let’s give a brief example to highlight the mechanisms we’re talking about. This basic understanding will help when we go on to discuss the issues in more detail further down in this article.

Imagine a manager is steering a portfolio of projects. A machine gives them information showing which projects are running well, which ones are going badly, and prioritising them. So far, so good. But what if the machine tells the manager that a project they believe in and are sponsoring is going to fail? The fact is, we humans tend to trust positive messages, but we’re quick to question negative predictions. In this situation, where the manager’s bonus is in jeopardy because their compensation is linked to their portfolio, they might be tempted to manipulate the algorithms or data to get outcomes that work. Naturally this skewing is bad for business.

So there are two sides to trust: trust that machines will provide reliable, unbiased information and predictions; and trust that humans will be act on the machine’s input without adding bias of their own. Naturally trust is good, but control is better. We’ll get back to that later.

How far is AI from eliminating human project managers?

First let’s take a look into our crystal ball. Research and advisory firm Gartner says that 80% of today’s project management tasks will be eliminated by 2030. While we at PwC think there’s some truth in this prediction, we don’t believe the changes will be that quick or radical. For one thing, applying artificial intelligence (AI) to project management is an incredibly complex challenge because it involves considering and reconciling the interests of opposing stakeholder groups.

That’s not to say progress isn’t being made. We’re currently at the stage of standardisation and automation, with machines able to fill in documents and create project charts that have been standardised. We’re also seeing more widespread use of chatbot AIs, which can tell a project manager or sponsor what’s going on, what the team is currently working on, and so on.

The next phase will be more predictions and recommendations on the basis of machine learning (ML) – the project management equivalent of finding out what the weather will be like tomorrow and deciding whether you should take the kids on a picnic or visit a museum. Once this phase comes to pass, it will be the first time that a human project manager can be eliminated, because if project sponsors want to know what stage things are at, they’ll be able to ask a machine. But we’re not there yet.

What’s the endpoint?

After that, the next development will be the fully automated project manager: the equivalent of a self-driving car. We think it will be at least 20 years before this happens. Why? Primarily because there’s still not enough structure in organisations’ project management environment and data −all of which has evolved organically.

The first step in the long journey to fully automated project management will be to prepare this environment so that it can provide the kind of structured data a machine needs to make predictions.

Another issue hampering progress on the path to automation is the complexity of the environment a machine has to keep track of to make decisions. A self-driving car basically only has to monitor the road ahead. A ‘self-driving’ project management machine, by contrast, will need to trawl the whole environment, all the time. At the moment, most of the relevant information is in the hands and heads of project managers. Some is stored as data but is not fully available in digital form in the project environment. Project managers understandably don’t have an interest in relinquishing their hold on this information, because it could make them obsolete.

What are the implications for project managers?

So one of the key questions is how to incentivise project managers to yield and adapt. The most obvious approach is to use AI to relieve project managers of the administrative burden and encourage them to concentrate on the strategic and people side of their role. Naturally not all PMs will want strategic work, and some will prefer to continue doing the admin even if a machine is capable of taking care of it. But on this point we tend to agree with Gartner’s prediction that classic project management tasks will be eliminated or automated in the next couple of years.

Obviously this will result in massive changes to project management teams. Chatbot assistants will be handling a lot of the routine questions human assistants used to answer, and bots will also be able to create documents. Companies will be investing heavily in this kind of technology because there are massive cost savings to be had.

The clear message for humans working in project management is therefore that they should be focusing on educating themselves and developing their people, change and stakeholder management skills. The stark fact is that project managers who only know the standard could soon be out of a job.

What are the implications for management?

But what are the messages for top management? One important point to realise is that however tempting cost savings might be, it’s important to strike a balance between cost reduction and quality reduction. Among other things you need a layer of quality control to challenge the outcomes created by the machine and check the documents it generates. This way you allow the machine to learn and get better and better.

Managers also have to think about how to build and preserve the trust we talked about earlier in this article: trust in machines, and trust in people to make good use of them.

One way of building this trust is to educate and certify project managers to use AI and interpret the predictions it generates. They need to understand the underlying algorithms and data, how these combine to make predictions, and the risks involved. Secondly, you need ethical standards to guide people in the use of AI and set limits on the decisions you’re going to allow it to make. You also need transparency on the methods and algorithms used by AIs to avoid program bias (the assumptions made when the software is written and the way the algorithm is fed, which can result in skewed predictions and decisions). An extra layer of governance can be put in place to make sure no one is able to manipulate the algorithms and data. Third-party assurance completes the process of building trust and taking the whole bias out of the equation. Just as automakers have rules governing the accountability of the manufacturer, programmer or operator of a robot arm in the event that it hurts somebody, but also have an inspector come by periodically to check the equipment, you need independent assurance that you have the right rules and processes in place and that they’re working properly.

An important point to remember is that despite all the hype around AI, few people really understand the true limitations and benefits. To assure that their investment pays off, companies need to understand the prerequisites for the successful application of AI in project management, negotiate the complex environment and social dynamics it involves −and keep a detailed record.

To move or not to move?

Then there’s the question of when to make a move and how far to go. Taking the leap into AI entails effort and risk, but the first company to create a data layer to allow machines to steer and make predictions on the basis of data (rather than continuing to operate with all the information in the heads of human project managers) will gain a huge first-mover advantage.

Smaller companies obviously don’t have an urgent need for AI in project management because they have the whole overview. But it’s a different matter for big organisations. Looking around the market, we see two industries with a particular interest in moving fast. The first is pharmaceuticals, who have a huge amount to gain from being able to reliably predict which projects in their R&D portfolio will end up as products – and as a consequence are investing heavily in AI technology. The second is financial services, where providers are under intense cost pressure and having to examine each investment closely and call an early halt early if the outcome isn’t going to justify the human and financial resources involved. Top managers at these companies see a lot of numbers, but often have no clue what’s behind them. AI could step into this breach to tell them what the figures mean.

Article overview