Hero Image

Leveraging LLMs as a Product Owner:

 

20 Jan 2025 | Melissa Bonsted

Leveraging LLMs as a Product Owner: When They Help, and When They Don’t

With AI becoming increasingly embedded in daily work, Product Owners have an opportunity to use it as a valuable assistant. As responsibilities grow, leveraging AI can help lighten the workload in key areas. This article, from experienced Product Director, Melissa Bonsted, explores how Large Language Models (LLMs) can assist Product Owners and highlights the situations where human expertise remains irreplaceable.

Discovery Phase

Product Owners often face the challenge of quickly ramping up on key details: who the users are, what their needs are, and why the product matters to them. In remote settings, this challenge is amplified due to limited direct interaction with users.

LLMs can be incredibly helpful during this phase. Before the first call or meeting, they can provide a high-level understanding of the client’s domain, typical user personas, and common pain points. In just 5-10 minutes, LLMs can help gather foundational knowledge, enabling more pointed, relevant questions during discussions.

In Practice: A Product Owner preparing for a last-minute client call could use an LLM (such as Chat GPT) with prompts like: “Explain the casino/hospitality domain with a focus for a product person to understand it. Highlight how loyalty rewards components drive customer growth.” Follow-up prompts could explore other relevant pain points. This preparation can ensure efficiency and informed participation in meetings.

Build Phase: Story Creation

When breaking down a new feature, deciding how to slice it and where to begin can be daunting. LLMs can assist by generating story starters based on brief descriptions of the feature.

These story starters act as seeds. While the LLM provides the initial draft, Product Owners refine them with domain knowledge, business context, and stakeholder insights. This process overcomes writer’s block and uncovers aspects that might have been overlooked.

In Practice: A Product Owner addressing a notes/comments feature could prompt an LLM with: “Describe a feature where users can view, add, and edit notes with various attributes. Suggest base user stories and related system areas for integration.” The LLM’s output could then be customized to fit specific requirements, saving significant time.

Build Phase: Feature Ideation

When a product feels stagnant, LLMs can act as creative partners for exploring new angles, brainstorming potential markets, or improving features. While they offer valuable sparks of innovation, LLMs cannot replace user feedback or collaboration with stakeholders and teams.

In Practice: A Product Owner struggling to get actionable user feedback might ask an LLM: “For users in the domain, what unique features or pain points could exist that aren’t being shared? How might these be addressed?” Reviewing the results could inspire ideas to investigate further with users, potentially leading to valuable enhancements.When to Be Cautious with LLMs

While LLMs are powerful, caution is necessary in certain areas:

  1. Use It Ethically: Ensure compliance with ethical guidelines, especially regarding user privacy and sensitive information.
  2. Maintain the Human Component and Empathy: Empathy remains essential. LLMs lack the emotional intelligence to deeply understand users, so prioritize human interaction for gathering insights.
  3. Be Aware of Bias: LLM responses may reflect inherent biases in their training data. Critically assess outputs to navigate these biases effectively.

Why Not Just Copy-Paste?

After creating a story or acceptance criteria, it might be tempting to copy-paste the output directly into a ticketing tool. However, LLMs lack contextual awareness of your environment, domain intricacies, or stakeholder nuances.

Use the LLM’s output as a structured starting point. Customize it to fit your specific context, ensuring that it accurately conveys what needs to be built and why it matters.

The Human Touch Matters

LLMs can provide surface-level understanding of users and domains, but they can’t replace real conversations. Users live their jobs daily and understand their pain points better than any AI model. Building strong partnerships with clients and translating user insights into actionable systems remains a hallmark of effective Product Ownership.

Be Cautious of Bias

The quality of LLM outputs depends heavily on the quality of prompts. Provide comprehensive context to achieve better results. Refine prompts, review outputs critically, and iterate as needed to maximize value.

Summary

LLMs such as Chat GPT an co pilot are here to stay and will become increasingly integrated into mainstream products. Product Owners should embrace their potential to enhance workflows while understanding their limitations. By balancing AI-driven insights with human empathy and domain expertise, better products can be built.

Remember, an LLM is not a human. It lacks the context, fluency, and logic that comes with human experience. However, when used thoughtfully, it can become a powerful assistant, helping Product Owners navigate their expanding responsibilities and focus on what truly matters—building great products.