I once helped to establish the tradition of tech talks in one company. To encourage engineers to make talks, I found the opposite advises to be useful: remove as much as possible the barriers to make presentation. It should take similar time to prepare presentation as it is to make it.
As an engineer, I find it inherently wrong to create content (here: create slides in Apple Keynote), then export this as an image to extract the content again via OCR afterwards. This could be more direct/better, no?
The main reason I use OCR here is that I often have slides which include screenshots of web pages or terminal sessions, so the slides themselves are image-based right from the moment I created them - there's no text to extract.