Post by Roy Weiss.
AI and Machine learning are hot topics these days.
Companies are using AI to obtain valuable insights from data and make their products better. From a Product Manager’s perspective, this is an excellent opportunity to adopt cutting-edge technologies to solve a problem. That means more problems can be solved more elegantly.
However, an AI-based product presents many challenges for the Product Manager, and it often requires the Product Manager to use different tools and technical knowledge as machine-learning, statistics, and probability on some level.
AI-products unique challenges
- First, the relationship with the Data scientist is different from the conventional R&D developer: Development times are offbeat, uncertainty is much more significant, and often there are difficulties in communicating the needs.
- The most critical factors in AI-based products, (more than the algorithms!) are the size and quality of the data. The Product Manager must collect the right Data with the right features that would give the best result later. This task should be prioritized very carefully.
- The communication throughout the organization is also different since people do not like black boxes. The PM must explain thoroughly, but in layman terms, how technology work for the executive, customers, and sales. Otherwise, it may cause over expectation on the one hand or skepticism on the other.
AI-products conventional challenges
- As with any product, the Product Manager must make sure that real problems are solved, real value is provided, and the customer is likely to pay for the company to address its problems. In the case of an AI-based product, there is sometimes a tendency to forget the ‘main thing’ and focus more on the technology and execution than in the problem space.
- Last but not least are the matrix for success that may look different in an AI-based product as probability and statistical terms may be introduced.
It appears that the role of the Product Manager is stretched when it comes to an AI-based product, and there is a strong need to come up with different methodologies for this kind of products.