Machine learning is already all over in the digital marketing landscape. It is most likely not something you will own using proprietary data (unless you are a massive corporation), but is definitely something you can take advantage of if you aren't already.
Something as simple as bidding to a target CPA in Facebook while serving multiple ads would constitute machine learning. Facebook will learn what CPM to bid for each ad to convert customers at your target cost. It will also understand which ad is most likely to drive conversions to your target market. Any remarketing campaigns you run would also be considered machine learning, by automatically targeting users who have gone to certain pages on your site with lists that update in realtime.
In the SEM space, bid optimization engines such as Optimine Impact use machine learning to manage complex bidding strategies and make adjustments to individual ads and keywords.
For B2B targeted campaigns specially, you can take advantage of Linkedin's machine learning by using their matched audience feature. Once you upload your email list, you can build lookalikes, made possible by LinkedIn's (and their partners') machine learning.
Ultimately, "Machine Learning" itself is not as important to success in marketing as the individual tools and tactics. My suggestion would be to simply keep up with the newest tools and tactics being used in the marketing landscape as machine learning will ultimate be woven into anything most of these companies put out moving forward. Continually test new options to see what works best for you.