As data collection and AI grows in demand, the availability of AI- & ML-services has commoditized and we are seeing a similar development than for cloud computing in the last decade. This implies that the success factor in applying AI for SaaS companies is not to built superior models, but rather to train available data in the best possible way so the models can suite the needs of the business context. This is a major shift in the role model of engineers from writing algorithms to training of applications and come with challenges of its own. Key success factor on this journey is to have training data at sufficient quantity and quality as well as further data sets to test the model prior to moving into production.
AI- & ML- algorithms are in general designed to serve on of the following purposes:
- Diagnose a situation / pattern recognition
- Predict a future outcome
- Prescribe actions in order to optimize along certain dimensions
- Personalized user experience
Based on their business model and customer value proposition, companies need to identify areas where algorithms have the potential to increase status quo by 10x and also determine what data is needed for this.