At a recent networking event, a product manager from a successful consumer app startup asked me about the key differences between building an AI product and a non-AI product. This inspired me to share insights from 15 years of experience as a Founder and in ‘Product’, with a focus on AI for the past 8 years.

Acquire technical skills. You should gain a solid understanding of AI technologies, machine learning algorithms, and data science principles. This knowledge is crucial for informed decisions and communication with technical teams. Without a PhD and with only a few hobby projects in 'Data Science,' I was able to make the move. I learned ‘on the job’ and ramped up my understanding by asking our ML/NLP scientists many questions.

Understand data management. Develop skills in managing and analysing large datasets, which are fundamental to AI product development. This includes understanding how to collect, clean, and utilise data ethically and effectively.

Learn continuous improvement and continue to use agile methodologies. Agile development practices won't be anything new to you. But make sure to focus on the specific parts that support rapid prototyping, testing, and iteration. These are key to the iterative nature of AI products, and can catch you by surprise. View the whole pipeline as iterative and understand how upstream data differences can impact the downstream product.

Navigate ethical and privacy concerns. I don’t need to write a lot, you see the news stories. Avoid being one. Address ethical considerations, data privacy, and biases in AI models. Understanding these aspects is vital for responsible AI products.

Collaborate across disciplines. Work closely with data scientists, engineers, and stakeholders. Effective collaboration ensures diverse perspectives are integrated into product development. In a large company, one of the key things to do is educate 'business stakeholders' about AI, brainstorming to build proof of concepts, and get feedback from users early and often. Make them part of the product team.

Engage in lifelong learning. Another obvious one, but as you can see from the initial release of ChatGPT to the version we have today, as well as all of the open source alternatives, this space moves rapidly. Stay updated with the latest advancements and trends in AI. Continuous learning is essential.

Adapt PM frameworks for AI. Utilise and adapt traditional product management frameworks for AI product development. Address AI challenges such as data quality and model interpretability.

Develop a vision for AI integration. Strategically plan how AI can be integrated into existing products or used to create new ones. This requires a clear understanding of AI capabilities and their potential impact on business outcomes.

Closing words: I'd recommend going through one of the many courses that are now online that will take you through, step-by-step, building a basic AI/ML product. A number of them also now have more advanced courses, if they align with your specific modelling, product or niche needs. It most likely goes without say, but I'll say it anyway — 'don't just try and square-peg round-hole an AI strategy or solution just because it's 'du jour'.