AI IQ

As Generative AI becomes more integrated into decision-making, from content generation to project insights, it is essential to understand how it arrives at its outputs. This requires emphasis on explainability, which refers to our ability to understand why and how an AI system makes a specific decision. Unlike traditional algorithms that follow clear, rule-based logic, Gen AI models such as ChatGPT operate based on patterns learned from vast amounts of data. These models use probabilities to predict the most likely next word, sentence, or outcome, based on context and training.

But this process can lack transparency. To address this, researchers and developers employ techniques such as attention mapping, input attribution, and model visualization to identify the factors influencing the AI’s outputs. For example, explainability tools can highlight which words in a prompt had the most influence on the model’s response or indicate how confident the model is in its predictions. Some of these capabilities are built into generative AI platforms, such as ChatGPT, through conversational prompting and token probabilities. Others, such as LIME and SHAP, are standalone Python libraries designed to interpret traditional machine learning models and provide insights into model behavior.

Why does this matter? In high-stakes environments such as healthcare, finance, or project management, stakeholders require transparency to trust and validate AI-driven insights. Explainability supports accountability, helps detect bias, and ensures that decisions align with ethical and legal standards. It bridges the gap between powerful AI capabilities and responsible human oversight, ensuring that Gen AI serves as a reliable partner in decision-making.

As generative AI (Gen AI) becomes increasingly embedded in project management processes, organizations must develop a clear and responsible framework for its use. Gen AI, such as ChatGPT, can create schedules, draft procurement strategies, assess risks, and even suggest team configurations. However, employees require guidelines to reduce risks that may undermine both project and organizational integrity. Adhering to ethical considerations when interacting with Gen AI requires a framework that aligns with the organization's ethical policies, reflecting its values, regulatory requirements, and fairness.  Here are three components to a good framework.

1. Governance. Governance is the foundation of an ethical Gen AI framework. It ensures that organizational use aligns with internal values, external regulations, and changing technology landscapes. Without strong governance, ethical breaches may occur without accountability or oversight.

Organization Responsibility:        

• Develop formal policies on AI usage
• Monitor evolving legal and ethical standards
• Establish audits and regular reviews
• Communicate responsibly and clearly

2. Training. Employees engaging with Gen AI must be equipped to use it responsibly. This includes understanding how prompts influence outcomes and ensuring proper validation and attribution of AI-generated content. Potential issues include unclear or manipulative prompts, failure to fact-check AI-generated output, ignoring attribution norms, and making poor decisions based on unverified content. Employees require training on ethical challenges to ensure they are accountable for their interactions with Gen AI.

Training Objectives:    

• Use transparent and honest prompts. Avoid prompts that intentionally mislead the system to generate biased or overly persuasive content
• Verify and validate outputs. Don’t assume Gen AI responses are accurate. Always ask for the sources and cross-check facts, numbers, and claims with trusted sources
• Respect attribution. If Gen AI uses or summarizes identifiable content, cite or acknowledge it appropriately
• Recognize limitations. Understand what Gen AI can and cannot do. Avoid delegating critical decisions solely to the tool
• Seek feedback. Share your Gen AI-generated work with peers or supervisors before taking action based on it

3. Data Management. Gen AI results are based on the quality of data it can access.  Datasets may be incomplete, inaccurate, or biased. Access may be given to confidential information or contain personal data without informed consent. Data ownership may be unclear, and sharing it could violate ethics policies.  

Data Imperatives:

• Consistently evaluate data for credibility and completeness
• Differentiate between access to internal data and external data, identifying risk for each
• Determine access restrictions for privacy, security, and the potential for misuse

The consequences of ethical violations when using Gen AI can be significant (Hagerty & Rubinov, 2019). Poor decisions and breaches of ethical policy can not only have serious repercussions but may go unnoticed for some time. Accountability is necessary from all parts of the organization involved. Employees must acknowledge their use of Gen AI and understand the associated level of risk. Ethical lapses in the use of Gen AI can also have cascading effects, from poor project outcomes to reputational damage. Accountability must be built into every layer: policy, practice, and personnel. Organizations should not only empower employees to utilize Gen AI but also equip them to be stewards of its ethical application.

The technology offers incredible opportunities for organizational efficiency, which must be balanced by defining and implementing a framework that upholds ethical standards.

Is your organization ready to integrate Gen AI responsibly?

Some references to check:

Dignum, V. (2019). Responsible artificial intelligence: How to develop and use AI in a responsible way. Springer. 

Hagerty, A., & Rubinov, I. (2019). Global AI ethics: A review of the social impacts and ethical implications of artificial intelligence. AI & Society, 36(1), 55–66

In today’s fast-paced, data-rich environments, project management is no longer just about tracking milestones or balancing scope, time, and budget. It’s about predicting outcomes, preventing failure, and optimizing performance. Machine learning (ML) is an opportunity to gain a competitive edge in project delivery. A machine learning model is developed by learning patterns from information and is used to make predictions or support decisions.

Building and using machine learning models in project management is a strategic opportunity that empowers teams to shift from reactive problem-solving to proactive decision-making. Models are reusable assets that grow more valuable over time as more data is added. Once integrated into project processes, models can be used to deliver fast, objective insights that help project teams and executives make better decisions with less cognitive bias.

Machine learning models can support project management in several ways:

1. Predictive Forecasting: ML models can estimate the likelihood of cost or schedule overruns, identify which projects are at risk, and flag potential bottlenecks before they occur.
2. Resource Optimization: By learning from previous projects, models can predict where resources are likely to be overcommitted or underused.
3. Risk Identification: By utilizing project characteristics and environmental factors, models can analyze risk, providing early detection and mitigation plans.
4. Decision Support: Instead of relying on biased intuition, project managers can receive insights based on data to justify contingency plans or identify decisions that require escalation.

There are three options for organizations that want to build project models. The first is internal-only models, which utilize project history, KPIs, and internal metrics. This approach is ideal for highly customized or confidential projects. The second is a hybrid model, combining internal data with publicly available datasets or third-party repositories to increase model generalizability and robustness. The final type involves using external models that are created and made available outside the organization, but with sufficient applicability. These can be useful for small organizations that rarely undertake projects or for any organization that lacks sufficient internal project data.

Machine learning is a powerful technology that elevates project management from hindsight to foresight. Organizations that invest in building or adopting ML models gain an advantage in delivering projects more accurately, efficiently, and confidently.

Artificial intelligence is no longer a distant trend—it's actively reshaping how project managers plan, monitor, and deliver results. From forecasting project risks to generating reports through natural language processing, AI is unlocking new efficiencies. For PMOs and project professionals, this isn’t just an evolution—it’s a transformation of the project landscape.

Yet, the most significant barrier to AI adoption isn’t the technology itself. It’s the readiness of the people expected to use it.

Project managers are uniquely positioned at the intersection of strategic oversight and operational detail. To lead AI-integrated projects, they must understand not only how AI works but also how to collaborate with it effectively. This means gaining literacy in tools like machine learning, process automation, and predictive analytics—not to become data scientists, but to confidently interpret results, assess model performance, and apply AI outputs to decision-making.

PMOs have a key role in fostering this shift. Developing structured AI learning pathways ensures that project teams are equipped for what’s ahead. These pathways should be role-based, scalable, and practical, covering everything from ethical AI usage and data management to real-world use cases in project environments. Importantly, they must recognize that AI does not replace the core competencies of project management—it enhances them.

A supportive community of practice and mentorship model can further accelerate adoption, turning training into shared experience. By embedding AI into the PMO learning culture, organizations can create a workforce that is agile, informed, and capable of leveraging AI as a strategic advantage.

AI skills development isn’t a side initiative. It’s essential for the future of project leadership.

Although AI is a powerful new technology, there will likely be missteps as we learn to use it. Technology users have been lulled into expectations that software will deliver results with minimal effort. Checking local weather on a smartphone, shopping online, or ordering a ride from Uber are easy tasks. For project management, software can schedule tasks and manage complex requirements. This can be accomplished without insight into how the software logic works. However, for AI, we need to step back and understand the process, sometimes questioning the results. AI is not a turnkey solution, yet we treat it like it is. Think about using a large language model (LLM) like ChatGPT. The user asks a question and receives a response. Most users do not think about how the response is generated or what data was used to produce the result.  To clarify a vague answer, a new prompt can be entered requesting more details.

As project managers, we need to know more. AI methods can include supervised learning, unsupervised learning, reinforcement learning, semi-supervised learning, self-supervised learning, and genetic algorithms. Each algorithm has a different process for calculating results and has different data requirements. AI prediction is a probability. When you receive a response from an LLM, do you understand the probability that the answer is correct? You can ask for the source to help validate the answer. 

The way out of this forest of possibilities is education and training. We don’t need to be data scientists or software engineers, but we have a responsibility to investigate and understand how the algorithm provides answers. There is a learning curve with AI technology, and business users can be trained to enhance their knowledge so they know how to acquire optimal results from AI technology.