You can’t exist in the tech world today without hearing about AI at every turn. The business value is clear— automate tasks that would take humans far longer and gain insight into valuable new data. Given that our product managers regularly use data to support their hypotheses about new features, where do they fit into this AI equation? How can they overcome any knowledge gaps and lead successful product launches utilizing machine learning?
The answer is not as complicated as it may seem. A product manager with a solid grasp for the tactical process can rise to the challenge.
We know that product managers need to think tactical, but how deep should their AI knowledge go? Ultimately, product managers just need to know what AI can do, and conversely, what the limitations are. This is important in determining whether AI is even a viable solution for the feature in question. Knowing what data is available, and how it can further AI development allows the product manager to tactically steer the product.
Taking it a step further, the product manager should be able to know how AI will integrate into the current product, and how to improve the AI models moving forward. Do the collected data aid future initiatives? What other data sets could be useful for future AI capabilities in the product? These are questions that can help propel efficiency in the machine learning model. The key is to not get bogged down with implementation level questions and stick to the high-level strategy.
A good exercise is to test your knowledge of useful applications of AI by thinking about how it can aid your own company. A good place to look is at your rule-based systems that depend on certain classifications. For example, here at Leapfrog, we came up with our management of employee supervised learning. This system depends on classifications such as whether people are high or low performers, their total length of employment, and employee feedback. We can use this to determine goalposts of what employees can achieve in 3 months, 6 months, or in one year. Continued data and feedback over time will grow these models. Ultimately, we can strengthen company culture, morale, and growth by investing AI into our employees. Looking at simple internal solutions allows us to start flexing our machine learning muscles.
A product manager can use their strategic knowledge to structure an AI project appropriately. The following high-level steps can guide the flow necessary for an AI project.
Get the data: Some clients may desire an AI solution, but will not have the tools in place for this to be viable. Without the right set of customer data, you will not be able to strategically apply it to an AI problem.
Explore and analyze the data: How can I use the data? Does this data support my product hypothesis? What initial findings do I see? The product manager can engage with the data at a high level to see if this fulfills the needs of the feature in question.
Validate the hypothesis with users: As with any project, it is imperative that we validate our assumptions with our user. Testing our initial ideas will allow us to make valuable iterations throughout the product cycle. By starting small and iterating, product development can gain speed.
Create quick and dirty models: The team can use the initial research and hypothesis to begin building quick models. More questions will arise, but don’t lose sight of the initial hypothesis and goal.
Tune the models: Iterating on the initial models will continue to confirm or deny the original hypothesis. We can ask ourselves, “What is the minimum level of functionality that is acceptable?” Keep tuning this model until solid results are delivered.
At the end of the day, the biggest measure of success will be if a product manager can ask the right questions at every stage of the product cycle. They are steering the features from hypothesis through development. Success in machine learning projects is ultimately measured similarly to regular development projects. If the product manager can retain their tactical approach, keep the team aligned around key product goals, and stay focused on the experience of the end-user, then the project will be in good hands.