I have a secret to tell you…
I’m not the sharpest tool in the shed and AI had me scared.
As a project manager I had no idea at the beginning of the year that I would need to know how machine learning and AI algorithms work.
Then, an AI project was assigned to me, and I needed to get up to speed on machine learning (ML) so I could understand the basics.
I learned that machine learning is just another tool in the PM toolbox—it helps project managers make educated decisions, streamline workflows, and give predictable results (and every project manager wants to do this).
Machine learning is like teaching a computer how to guess things correctly by showing many examples. Imagine you show it pictures of cats and dogs, then it learns to differentiate between them.
If you're assigned an AI project and understand how machine learning works, you will have more confidence in managing it and will not think you're in over your head.
Understanding the foundations of machine learning can help you become more valuable to your team and give you a valuable perspective on how AI actually works.
I'm going to try to keep the discussion about ML and algorithms simple since it can get very technical very quickly.
If you stick with me, you'll have some knowledge you can apply to your next AI project or make you look like a star in front of your team.
The blueprint the AI follows
Let's start with defining what an algorithm is.
To put it in terms that every project manager would understand, algorithms are like your project charter. The traditional charter outlines the milestones and dependencies of a project but provides instructions on how the project should proceed. An ML algorithm provides instructions on how to take your data and turn it into automated actions or insights. Think of this as a project charter that learns and updates itself automatically as the project progresses.
An algorithm is a set of instructions, like a recipe, that tells a computer how to solve a problem or do a task.
A typical AI project may be a customer service system. Behind this system is an algorithm, which is a flexible set of instructions that allows the system to learn from past interactions with customers and apply that learning while responding to new customer questions. As I mentioned in past articles, you don't need to learn how to code the algorithm; you just need to understand how it works so you can talk to your team and stakeholders.
The model is the deliverable you deploy
You've probably heard references to "the model" when working with AI.
The model is generated after the algorithm is trained on your data. Just like a project deliverable is the end result of all your planning and coordination, the model is the deliverable you can put into action after it has been trained to recognize patterns in the data.
An AI model is like a smart robot that learns to do tasks or make decisions by examining many examples.
Maybe you're a project manager for a global company that makes widgets and you need to build an AI tool that forecasts demand to help even our production. The model in the tool will examine past sales worldwide to see when the demand for your widgets will be in the future. A model is a representation of patterns in historical data, and if you can understand that, you can talk to your data scientists to ask it the right questions, and you know its limitations.
Remember that since the model relies on accurate data—if you put garbage in, you'll get garbage out.
Train that model to make it’s own decisions
As a project manager, you just need a 10,000-foot view of how models make decisions.
In simple terms, a model is a set of interconnected decision points that looks like a complex decision tree. Each model makes decisions differently. In a neural network model, each point assigns “weights” to factors and prioritizes some links over others.
Training is when you teach an AI model by showing it lots of examples so it can learn to make predictions or recognize patterns.
Stakeholders will give you feedback on which features are the most important. Your AI model will allow you to ask smart questions over these features to generate its decisions.
You can let your stakeholders know these weights can be adjusted over time, so the system stays fluid as the conditions and priorities change.
Choosing the right learning approach
Everyone learns from their past projects, and you may even have “lessons learned” filed for someone in the future to review.
When your model "learns" from past data this is actually the phase where it is training. The training data determines the model's reliability and accuracy. Now you can see why having the best data quality is an important factor—you're not training your model correctly on out of date or poor data.
Parameters are like the little knobs and dials an AI model adjusts while it learns from the examples during training.
You can work with your data scientists on your team to source training data that reflects the types of projects your organization has, which can make better predictions for your projects.
Most projects require a customized approach, and your machine learning project also learns faster with the right learning method. Three types of learning are commonly used.
Supervised Learning is like a tightly controlled project with clear milestones. The model learns from labeled data—which are examples of the “correct answers,”—to help it improve its predictions. Take our customer feedback example earlier—this type of learning works well since you’ll have labeled examples (positive or negative feedback) to guide the model.
Unsupervised Learning is like a brainstorming session, but it's more exploratory. This type of learning looks for patterns in the data and doesn't use labeled data. Unsupervised learning can identify clusters and reveal hidden relationships without having pre-defined categories.
Reinforcement Learning is more like running an agile project. The model learns by trial and error, continuously improving over time based on feedback it receives and iterating. Reinforcement learning helps AI-driven chatbots adapt and improve, making them more effective as more users use them.
Lay a solid foundation with good data
Machine learning depends on accurate, consistent, and relevant data.
This is critical when your AI project relies on data from multiple sources, like customer feedback or CRM data. If you get bad data, you'll have a bad model. Nobody wants this since it’s unusable.
As a project manager, you should work with your team to clean the data and establish clear data governance protocols to ensure good data quality. You'll work closely with your data engineers to ensure data quality.
Setting this up correctly gives your model and project the best chance of success.
Build your model for the real world
Training and good data can only get you so far—you must send your AI model out into the wild.
It will see new data that it has never seen before. The model's ability to deal with this data and perform well is called generalization. For any AI project, you need to be sure the model can handle being kicked out of the warm training environment and into the cold, cruel world.
Model deployment is also known as model operationalization.
As a project manager, you should petition for diverse training data and ensure the model is rigorously tested before deployment. As with anything that is hammered on before release, it reduces the risk of unexpected failures when it's out in the field.
Final thoughts
A little knowledge can go a long way in helping you push for organizational AI projects, manage stakeholder expectations, and make data-driven decisions and not guesswork a part of your projects.
AI projects will start becoming more frequent for most of us in 2025, and getting comfortable with these concepts will empower you to leverage AI strategically in your job and organization. You’ll be able to set better KPIs and communicate more effectively with technical teams, and you can even enable methodologies like CPMAI to help you conquer AI projects.
So, don’t dread working with AI and machine learning—it’s your next ally in achieving smarter, more efficient project outcomes.
AI-Driven Tools for PMs
Sembly AI - AI meeting notes, tasks, and more—sent automatically to your favorite apps.
FigFlow - Translate designs into user stories.
OneNode - An AI-powered note-taking app.
Feta - Tool for efficient meetings and automated post-meeting tasks.
Notaic - Use LLM agents to answer emails in your voice.
AI News PMs Can Use
The Power of Artificial Intelligence for Project Management
The Bain “Labor 2030” Report projects U.S. companies to invest $8 trillion into automation by 2030.
According to the Census Bureau report, 96.2% of U.S. companies haven't even started implementing AI.
Cool ChatGPT Prompt for PMs
Project Methodology
Design a hybrid project management methodology that combines elements of Agile and traditional approaches for complex, long-term projects.