What AI Is and Is Not
Artificial intelligence refers to systems that perform tasks requiring pattern recognition, prediction, generation, or decision support.
For product managers, the useful distinction is not science fiction versus reality. It is:
<ul>
<li>narrow AI: built for specific tasks or classes of tasks</li>
<li>general intelligence: a theoretical concept that is not how most products are built today</li>
</ul>
Most real-world AI product work is about narrow, useful systems.

Which system is the clearest example of narrow AI?

Most AI products PMs work on today are narrow AI systems built for specific tasks.

An AI Product Is More Than a Model
A usable AI feature usually combines several parts:
<ul>
<li>the model</li>
<li>instructions or prompts</li>
<li>context or retrieval</li>
<li>tools or integrations</li>
<li>UX for review and recovery</li>
<li>monitoring and evaluation</li>
</ul>
PMs get into trouble when they treat the base model as the whole product.

A model keeps giving generic answers to internal policy questions because it has no access to the company handbook. What is most obviously missing?

A great base model alone guarantees a good AI product experience.

The PM Lens on AI
Before adding AI, ask:
<ul>
<li>What user task or decision improves?</li>
<li>How much error can this workflow tolerate?</li>
<li>What happens when the model is wrong?</li>
<li>What will success look like in product terms?</li>
</ul>
Those questions matter more than whether a feature sounds advanced.

Which question is most useful before adding AI to a product?

PMs need to understand AI capability boundaries even if they never train a model themselves.