Integrating with existing AAC.
Multimodal communication is the field's own best practice. Treat it that way.
A learner who uses a speech-generating device for fast functional requests and a letterboard or a text-to-speech iPad for generative spelling or expanded communication is using AAC the way the AAC literature has long recommended: layered, contextual, multimodal. The pretense that text-based assisted communication has to displace existing systems is a false choice.
Where SLPs add the most value is in working together with the individual and the team to provide total communication supports — helping the person communicate in whatever method works for them, supporting vocalization if able, and assisting with regulation across the whole communication ecosystem — not in policing which tool is the "real" one.