AI for Non-Technical Project Managers: A Plain English Guide

If you've been handed an AI project to manage and you're not sure what half the words mean, this post is for you. I'm not going to explain the maths behind machine learning. I'm going to explain the things you actually need to know to do your job.

What "training a model" means

A machine learning model learns from examples. You give it a lot of historical data — say, a million past customer orders — and it finds patterns. Then it uses those patterns to make predictions on new data it hasn't seen before.

As a PM, you need to know two things about this process. First, it takes time — often weeks. Second, the quality of the output depends entirely on the quality of the input data. Garbage in, garbage out is not a cliché; it's the most common failure mode in AI projects.

What "accuracy" means and why it's not simple

When your data science team says the model is "90% accurate", ask: accurate at what, exactly? There's a big difference between being right 90% of the time on easy cases versus being right on the hard ones that actually matter to the business.

Ask your team to show you accuracy on the types of cases your stakeholders care about most. That's the number that matters.

The three questions to ask at every AI project review

1. Is the training data still representative? The world changes. A model trained on data from two years ago might be learning patterns that no longer apply.
2. How are we monitoring the model in production? After launch, someone needs to watch for "model drift" — when predictions start getting less accurate over time.
3. What's the fallback if the model fails? AI systems go wrong. There should always be a manual process or a simpler rule-based system that kicks in if the model has to be taken offline.

How to talk to data scientists

Data scientists often speak in technical terms not because they're trying to confuse you, but because precision matters in their work. The best thing you can do is ask simple questions without apology: "Can you explain that in terms of what the user actually sees?" or "What does that mean for our go-live date?"

They'll respect the directness. And you'll make better decisions with a clearer picture.

The most important thing

You don't need to be a technical expert to run a successful AI project. You need to be a good project manager who asks the right questions, manages expectations honestly, and keeps the human impact of the product front and centre. That's still very much a human skill.