OHT London x Product Managers: AI Sanctuary

It was a sunny autumn day as 18 professionals from diverse backgrounds, including commercial sectors, business development, product management, research, and startups gathered at the OHT AI Sanctuary session. The discussion, led by Mary Yip and Charly Massey, focused on the critical role of human-centred skills in AI product management. The session followed Chatham House rules, fostering an open and candid dialogue.

The conversation began by acknowledging the post-AI hype era and the increasing necessity to build trust in AI technologies. Trust in AI technologies covers not only the high-quality, representative data it’s built on but also how people want to interact and use AI technology. We’ve summarised some of the key topics below.

Understanding your users

It's important to explain AI products in a way that everyone can understand, including the people who need to make decisions about using them in their organisations. 

Get real

Seek candid feedback and be willing to question the necessity of features in development. As Product Managers, we balance the needs of super users with those of general users, it’s easy to over-cater to the most vocal users at the expense of broader usability.

Addressing Workforce Concerns

Acknowledge the widespread concern about AI's potential impact on jobs. By designing AI tools with users in mind, we can show how they can improve healthcare for everyone, from doctors and nurses to patients. 

Market readiness

The rising costs of AI development and deployment are a significant concern, with some organisations adopting a wait-and-see approach in anticipation of future cost reductions. The need to balance current needs with future developments while maintaining consistency was highlighted as a key challenge for product managers.

Reliability and Responsibility in AI Applications

Large Language Model (LLM) aren’t always perfect. They can make mistakes or perform differently in different situations. One idea is to use them for specific tasks or patient groups, similar to how doctors prescribe medicine for certain conditions. This would help to ensure they're used safely and effectively. But this raises a big question: who's responsible when AI tools are used in healthcare? This is especially tricky when they're used for things they weren't originally designed for. Software engineers and data scientists might feel uncomfortable with this uncertainty and supporting your team is another necessary human skill.

AI isn’t a standalone solution 

While recognising AI's potential to alleviate administrative burdens and improve patient adherence, human involvement is still important in care delivery. Some patients’ acceptability of AI varies depending on its application, with administrative tasks being more readily accepted than clinical decision-making. We talked about how combining AI strengths with human involvement were examples that could benefit areas such as diabetes management and social prescribing.

The AI Education Landscape

AI education of users varies widely. How do you design for this? There’s a need to separate human-AI interaction as distinct from traditional human-computer interaction. How do you ensure users (such as clinicians) can interpret AI output effectively as part of clinical decision-making? There’s been limited research into the interpretation of explainable AI metrics so far along with user design.

AI Tools for Product Managers

What AI tools are used by product managers themselves to enhance their work? Effective use of AI tools, particularly in areas like prompt engineering, requires practice and skill development. The group took this question offline, sharing best practices and tips within the community to the Product Managers Specialist Network Slack channel.

Future Directions - what is next?

The session touched briefly on important topics that merit further exploration, including cybersecurity, data governance, and AI governance in healthcare. These areas were identified as potential subjects for future AI Sanctuary discussions, reflecting the ongoing need for in-depth conversations about the challenges and opportunities presented by AI in healthcare.

This OHT AI Sanctuary session fully explored the human-centred aspects of AI product management in healthcare. AI isn't just about tech hype; it's about understanding patients, doctors, and everyone else involved. We covered everything from ethical dilemmas to practical challenges. It's clear that as AI gets more and more powerful, these discussions are going to be essential to make sure it's used in a way that actually helps people.