What data assets do you have? What can you do with this data? How do you store all the data you have?
Having cataloged and stored your data you can start ideating through the business problems you can solve and decide if AI application is appropriate to work with the dataset you possess.
Security is still a challenge for AI/ML, specifically when it comes to swarm intelligence and Generative adversarial networks (GANs). If someone has ML software installed, he is vulnerable to one of possible attacks that can poison your clean data. When you store hundreds of gigabytes of data, discovering what pieces have been poisoned is a challenging task.
For all application of AI/ML not necessarily billions of data points needed, it can be done with even 500 data points and as AI field is progressing, there are a lot of pre-trained models which can be adapted with a few number of data points.
Data quality is a very expensive problem, therefore combining human abilities with ML by keeping human mind in the loop allows companies to benefit from much smaller date sets.
In this video we discuss general concerns over data quality with Jonathan Rivers (CTO of Fortunes Publishing), Felix Hovsepian (Professor of AI at Warwick) and Varun Aggrawall (CoFounder of Aspiring Minds)