Brain Power

Machine Learning Is Just Starting to Transform Decision Making—And Careers




With too much data and not enough talent, it's only a matter of time before organizations jump on artificial intelligence (AI) to turn all of the information they have into better decision making. Machine learning has the potential to help project teams process massive amounts of data to reveal project patterns, identify risks and predict outcomes.

Over the next three years, project professionals expect the proportion of projects they manage using AI will increase from 23 to 37 percent, according to a 2019 PMI Pulse of the Profession® report. So while wholesale AI-assisted project teams are still on the horizon, project managers who get a head start on understanding the power—and limits—of AI will have a professional advantage.

Three project professionals immersed in AI's power demystify the role of machine learning in project management decision making and how it can transform careers:


Audrius Zujus,
founder and CEO, Aresi Labs, software and digital developer, Vilnius, Lithuania


Geetha Gopal, PMP,
assistant manager, product owner, bot services and projects, Daimler South East Asia Pte Ltd., Singapore


Bruno Rafael de Carvalho Santos, CAPM, PMP,
project manager, Sedimentary Geology Laboratory and Coppetec Foundation, supporting the Federal University of Rio de Janeiro, Rio de Janeiro, Brazil

There's been a lot of noise about AI and machine learning changing project management. What's real?

Mr. Zujus: AI has been misunderstood for a long time. It's been so hyped that most people think it's now in a really advanced state and can do anything. But the reality is: Few people are actually using machine learning to help them on projects. It's very early for machine learning to be used in this application.

Mr. Santos: I agree. People misuse the terms “AI” or “machine learning” as if they're magic that can do anything we want. They can't. They're essentially analytical statistics that have their own limitations on how and where to be used. And analytical statistics has been used for a long time, for example, with risk management. Now it just has a different name: machine learning.

Ms. Gopal: It's gaining attention because of the availability of congruent data and simplified integrating technologies like application programming interfaces that facilitate intelligent machine learning and data-driven decisions. That ability to harvest, understand and use data makes machine learning valuable for project management—today and in the future.

What recent projects have you worked on where machine learning helped with decision making?

Mr. Zujus: We built a system to take in data from product descriptions to learn how to generate new product descriptions. We used natural language processing to create product descriptions from manuals, pictures and other data. Our team had to estimate how long the project would take, so we used a simulation to identify some bottlenecks between the design team and development team.

Mr. Santos: In our university laboratory projects, we have plenty of data on individual and collective payroll data, which allows us to use regression to estimate the human resources costs for the projects. These estimations are usually better than the traditional calculation, and our projects have been trouble-free with this since 2015.

Ms. Gopal: We have built and trained some neural net algorithms for intelligent predictions based on historical data which also learns new behavior and self-optimizes its decisions from the latest data. This helps us to proactively notify our teams on potential risks so they can be addressed. We have extended this capability to intelligent routing, alerting and guided conversations, and it's helped us improve handling times and service levels.

How can machine learning tools help project managers most?

Mr. Zujus: AI works best with a large number of data. And when I say large, I mean hundreds of thousands of data points, or at least tens of thousands. AI can then derive patterns from that data that we can use to make better-educated estimates, like whether you'll finish a project on time. This is already being done with Monte Carlo simulations, a statistical method that looks at best-and worst-case scenarios.



Ms. Gopal: I agree—data is the launch pad for organizations that seek tangible benefits from the AI wave. But the current capabilities are limited to project assistance rather than entirely intelligent and self-learning tools for project management.

Mr. Santos: Agile teams have limitations with AI too. Providing a lot of information to machine learning software can be difficult for agile teams. They don't have time for that between sprints. Or the data might be unstructured, making it difficult to parse into a traditional database.

Ms. Gopal: Ultimately, data-driven decision making will help project managers to look beyond common intuitive biases.

Mr. Zujus: For now, organizations that have access to well-prepared and clean large data sets will have a much stronger competitive advantage over those that don't. They probably will be the first ones to empower their project managers to make the most of AI, like when it comes to estimating project schedules.

How can machine learning help project managers with resourcing needs?

Ms. Gopal: It has the potential to reduce manual efforts and free up resources for more value-add activities. Risk monitoring, capacity management, communications, self-service of project data and project performance on-demand are some of the low-hanging fruits to begin with. The resulting freed-up resources can be routed to support other projects or focus on critical areas and work more closely with stakeholders.

Mr. Zujus: It could help with hiring, too. A company might build an interesting tool to find great candidates on LinkedIn, and it might use natural language processing to understand what's written on their profiles.

Mr. Santos: But there's a risk with resourcing. We need to pay close attention to how we create and use these tools and algorithms, or we can get into ethical problems. For instance, an organization might use machine learning to be less biased and more inclusive in its hiring. But if there's bias in the data sets, then machine learning will just perpetuate the bias.

How can project managers help ensure these tools filter the right data?

Mr. Zujus: Working with the data preparation teams would help a lot. Project managers underestimate how much work goes into the cleaning and preparing of data. Categorizing and sorting the data can take 60 to 70 percent of the time it takes to build an application.



Mr. Santos: That's so true. Machine learning is only as good as the data it's provided: Garbage in, garbage out. When I work with geologists and biologists here at the university, they don't think about how the data they collect can be used for statistical analysis. So there's a need to train project teams on data literacy—how to use and manage machine learning software. If project teams and data analysts don't communicate properly, their interactions will be fruitless.

Ms. Gopal: That's right. Ideally, those responsible for back-end systems from where data is consumed should be accountable for data quality of dependent systems. This will help project managers with their main concern: having a clean and aggregated view for project performance with relevant and real-time data.

How can machine learning help analyze past projects to identify and assess risks?

Mr. Zujus: An organization can put its project reports into a statistical or AI model to identify the most common risks and causes of failure. The same goes for lessons learned and project successes. AI can help teams look for patterns as an indication of what might happen next.

Mr. Santos: Risk management is an area where machine learning has already been used, and I think it will be one of the project management areas that will most benefit from machine learning. But in order to have better risk analysis, we need correct project information. Knowledge management is critical.

Ms. Gopal: To build a robust AI system, we have to consider all the data available, train the algorithms with as many data points and layers as possible, review the results and adjust data points until close to 100 percent accuracy. In the context of project management and risks, the first step for organizations could be to ensure unbiased documentation of all project-related data, from all projects across a portfolio or the organization itself, if possible.

How do you balance the benefits of machine learning versus your own project management instincts?

Mr. Santos: You need to trust the machine—but you also have to understand the machine. If you don't understand the data and the algorithms, you'll misunderstand the data analysis, and it will be useless. But with human interaction or stakeholder management, you have to trust your gut. Machine learning won't help you.

Ms. Gopal: I like to segregate decision items between critical and mundane. Critical could be risk mitigation steps or budget increases, while mundane could be follow-up tasks or triggering of workflows based on milestone progress. It would be safer to use machine learning outcomes as recommendations or guided decisions for critical areas with intuitive assessment of the recommendations.

What do you say to those who worry machine learning might replace project managers?

Ms. Gopal: History shows that those fears aren't warranted. For example, I've read studies that prove that the introduction of various household machines in the 1980s and 1990s helped bring more women into the workforce. Machine learning tools and automation will free people from tedious, repetitive activities so they can focus on strategic activities. Adapting to and embracing the AI wave—and harvesting the capabilities it can offer—could be a boon to project managers.

Mr. Zujus: And even in a more automated world, process-driven jobs like project management will have a bright future. Project managers will just have to learn to better interpret the meanings of data.

Mr. Santos: We all agree: Machine learning will not replace project managers. Project managers manage people, and machine learning can automate and accelerate some tasks, but it cannot manage people. It cannot substitute human interaction. PM

AI Anticipation

Artificial intelligence (AI) has the power to transform organizations—but only if their teams are ready to embrace change.



37% Proportion of projects that are expected to use AI in three years—up from 23 percent today


AI technologies that are having the greatest influence on organizations today—and how much their impact will increase in the future:



AI innovators follow five key principles* that the most visionary organizations apply to their AI investments. AI laggards have embraced none of these traits.


*Framework based on research by Accenture


Organizations that embrace AI have a clear advantage.


Sources: Pulse of the Profession®: AI Innovators: Cracking the Code of Project Performance, PMI, 2019; Worldwide Semiannual AI Systems Spending Guide, IDC, 2019



Related Content

  • Thought Leadership Series

    Medir lo importante member content open

    Este segundo de una serie de informes con PwC examina cómo el 10 por ciento superior ha aumentado el número y la variedad de métricas, más allá de los parámetros tradicionales de alcance, cronograma…

  • Thought Leadership Series

    Mesurer ce qui compte member content open

    Ce deuxième d'une série de rapports avec PwC examine comment les 10 % les plus performants ont augmenté le nombre et la variété de mesures, au-delà des paramètres traditionnels de portée, de…

  • Thought Leadership Series

    Medindo o que importa member content open

    Este segundo de uma série de relatórios com a PwC examina como os 10% principais aumentaram o número e a variedade de métricas, além dos parâmetros tradicionais de escopo, cronograma e orçamento,…

  • Thought Leadership Series

    Jūyō jikō no sokutei member content open

    PwC to no ichiren no repōto no kono 2-banmede wa, toppu 10-pāsento ga jūrai no han'i, sukejūru, oyobi yosan paramētā o koete, metorikku no kazu to tayō-sei o dono yō ni fuyashita ka o shirabe,…

  • Thought Leadership Series

    Messen, worauf es ankommt member content open

    Dieser zweite einer Reihe von Berichten mit PwC untersucht, wie die Top 10 Prozent die Anzahl und Vielfalt der Metriken über den traditionellen Umfang, den Zeitplan und die Budgetparameter hinaus…