Developing an AI product: 30 red flags to watch out!

85% of AI projects never get to deployment! From the projects that get deployed, only 60% ever deliver on their promised impact!

Image Credit: https://unsplash.com/photos/4y8A6Ve-3GE

In building AI products, the following are key red flags I look out for during team updates. In my experience, addressing these red flags proactively can significantly improve the success of your AI product!

1. Vague success metrics such as “we are building an ML model to increase customer happiness” How do you define “happiness?” Instead, focus on the most paired down metric that is measurable and sensitive to the desired outcome. An intermediate or proxy metric to happiness can be time spent to accomplish a repeating workflow (such as creating an invoice in an accounting software) or the number of times the referral link is shared.
2. “We are not clear how the model will be used within the existing product workflow” This is very common. More often we focus on going from data to insights, but miss out on the last mile from insight to the outcome. Simply predicting customer churn is of little value till is applied to the customer success process to proactively reach out and manage these customers. Delivering a well-tuned robust ML model that can be deployed in production can range from 6–24 months depending on the complexity. Being clear about the strategic value of the project is…