It's not worth trying machine learning projects unless you have a huge data set.
True or false?
Smaller companies are afraid to add machine learning features to their projects unless they have big data. They think that since they're not Amazon or Microsoft, they don't have a large enough data set to be successful in taking on machine learning projects or features.
There are definitely applications of machine learning that can work even on small data sets. Perhaps you can start small and prove out a concept, before investing in getting more data to build a larger model. You can also use off-the-shelf models in AWS, Azure, and GCP to solve generic problems.
With smaller data sets you will encounter problems that you need to be weary of, such as over fitting, bias, and data imbalance. With the right tools and people, there are strategies to overcome these problems.
Like any good project management program, you can invest in small wins to either prove out the concept. This can help you gain credibility in a new program, and then getting funding for bigger and bolder bets.