AI in Product Management: 

Key Takeaways from Our Recent Breakfast Session

Last month, we hosted a breakfast session with Product Management specialists to explore how AI is being used in practice across asset management. The discussion moved beyond the hype and focused on the reality for Product teams: where AI is already adding value, where adoption remains challenging, and what needs to happen for firms to make better use of emerging capabilities. 

1. AI is becoming part of everyday productivity. Personal efficiency tools such as Microsoft Copilot and ChatGPT are now widely used, particularly for drafting, summarising and organising information. However, participants noted that more complex use cases, such as market research, competitor analysis and interpretation of regulatory material, remain less reliable. Challenges include restricted access to sources behind firewalls, difficulty handling complex PDFs and inconsistent interpretation of detailed documentation. 

2. Good AI depends on good data. Internally built AI tools are often limited by the quality and structure of the underlying data. Large Language Models and agent-based solutions can create efficiencies where enterprise data is well governed, but they are not a substitute for strong data management. Several attendees highlighted that building verified, well-organised and AI-ready data infrastructure is a critical success factor. 

3. Access, training and practical support are still uneven. While most organisations are encouraging the use of AI, access to tools beyond basic personal productivity applications varies significantly. Participants described frustration with limited training and a lack of team-specific implementation support. Even in larger firms with dedicated AI teams, guidance was often seen as too general to help Product teams embed AI into day-to-day processes. 

4. The clearest benefits are currently administrative. Product teams are seeing useful efficiencies in areas such as note-taking, meeting summaries and preparation of Board or Committee papers. However, AI is not yet widely embedded in core delivery processes such as fund launches, share class launches or governance activities such as Assessment of Value. A common barrier was that teams are often too busy managing day-to-day delivery to step back and redesign processes around AI. 

5. Human judgement remains essential. The group agreed that prompt-based tools can provide a helpful starting point, particularly when building an initial understanding of a topic. However, outputs still require expert review, especially where accuracy, nuance and regulatory interpretation matter. Better prompting, including the use of truth protocols to reduce hallucinations, can help, but AI still has a greater “black box” element than many traditional technologies. 

6. Agentic AI could shift the opportunity from productivity to process automation. Participants recognised the potential of agentic AI, particularly for workflows with clearly defined steps, inputs and outcomes. While constantly changing capabilities make it difficult to create consistent AI-generated processes today, the longer-term opportunity may lie less in individual productivity gains and more in automating repeatable Product Management workflows. 

Thank you to everyone who joined the discussion and shared their experiences so openly. If you are exploring how AI could be applied more practically within Product Management, we would be happy to continue the conversation.