PM.

CommunityGlossaryBlog

AI for Product Managers

Cover Image for AI for Product Managers
Me
Me

AI for Product Managers: A Practical Guide

Is there such a thing as an “AI Product Manager”?

Let’s clear up a common misunderstanding: there isn’t really such a thing as an “AI Product Manager.” While the term is trendy and has sparked a lot of discussion online, it’s misleading. Product Managers (PMs) focus on the overall success of a product, while AI or Machine Learning (ML) Engineers handle building and improving AI models.

Most PMs won’t develop AI models themselves. Instead, they’ll use existing ones, like OpenAI’s GPT, Meta’s LLAMA, Google’s Gemini, or Anthropic’s Claude. The heavy lifting of creating and maintaining these models is done by engineering teams.

If anything, the term "AI Product Manager" might better fit future tools, like virtual PM assistants that can help identify customer needs, align them with business goals, and break them into tasks. While this idea is exciting, it’s still a bit futuristic and far from replacing human PMs today.

What stays the same for PMs in the age of AI

AI hasn’t changed the basics of product management. (I've shared my thoughts on how I think AI will change the job for PMs here. It's been a year ago, but so far it has aged reasonably well.) As a PM, you’re still responsible for:

  1. Understanding user problems: Knowing what issues users want to solve.
  2. Creating solutions: Coming up with ways to address these problems with your team.
  3. Defining clear tasks: Breaking solutions into steps your team can execute.

What’s new is the need to understand AI concepts so you can make better decisions and communicate effectively with your team.

The basic LLM terminology for PMs

You don’t need to be a technical expert to work with AI, but knowing some key terms helps:

  • Token: AI models break text into smaller pieces called tokens. A token can be a word, part of a word, or punctuation.
  • Context window: This is the amount of information (measured in tokens) the AI can process at once, including both your input and the AI’s response.
  • System prompt: Instructions that tell the AI how to behave or respond.
  • User prompt: The text you enter to interact with the AI.
  • Temperature: Controls how creative or random the AI’s responses are. A low temperature gives straightforward answers, while a higher temperature allows more variety.

Other settings, like frequency penalty or presence penalty, fine-tune the AI’s behavior, but you likely won’t need them in most cases.

How LLMs work on a high level

Large language models (LLMs) predict what comes next in a sentence based on the text they’ve seen so far. They’re trained on massive amounts of data and can mimic meaningful conversations, summarize information, or generate content.

For PMs, it’s important to know that LLMs are great at tasks like text generation and summarization but may struggle with complex reasoning or up-to-date real-world knowledge.

Learn prompting as a PM

Writing good prompts is like good communication—clear, concise, and to the point. As a PM, you’ll often need to write prompts to test ideas or generate content quickly.

Tips for writing prompts:

  1. Be specific: Clearly explain what you want the AI to do, including the format or tone.
  2. Add context: Provide enough background to guide the AI.
  3. Refine and iterate: Adjust your prompt based on the AI’s output to get better results.

Prompting is a skill worth practicing since it’s becoming a common way to interact with AI tools.
If your product has an AI-based feature, from my experience, you as a Product Manager will be involved in creating the prompts in one form or another. Most likely, you will be assumed to be one of, if not the best communicator on your team. Therefore, you’ll be an obvious choice to provide initial instructions to the AI and might also be involved in the fine-tuning process.

Practical tips for PMs

  1. Experiment: Use tools like ChatGPT, Claude, or Bard to get hands-on experience.
  2. Work with engineers: Collaborate closely with technical teams to connect product goals with technical possibilities.
  3. Stay updated: Follow AI trends because things change quickly. A good source to stay up to date is Hacker News — the most relevant changes often make it to the front page.
  4. Focus on value: Look for areas where AI can improve your product, like saving time, enhancing user experiences, or driving new revenue. Don’t make the mistake of adding AI just for the sake of adding AI.

Conclusion

AI hasn’t changed the core of product management—it’s still about solving user problems. But it does give you powerful new tools. By understanding what AI can and can’t do, you can make smarter decisions and help your team succeed.

Keep learning, stay curious, and remember: AI is here to help you, not replace you. Not yet at least :)