Speech-language pathologist / speech and language therapist — tasks, one by one

Assessing communication and swallowing

Machines are starting to judge parts of speech, but the assessment stays with therapists: in a small study an AI's judgements of children's practice attempts largely matched a clinician's, and no record found shows AI diagnosing communication or swallowing disorders in practice.

One small study and the absence of a record; it does not show how assessments are done in clinics.

Delivering therapy and practice

Software takes on the repetitive practice while therapists set it up: in the Big CACTUS trial, self-managed computer therapy improved word finding 16.2% more than usual care but not conversation; England's guideline recommends considering it alongside face-to-face therapy rather than instead of it; and in a small study AI ran children's drill practice after a clinician led the preparation.

Trials of practice for word finding and one speech sound; they do not show software leading therapy, and the guideline says the NHS does not routinely use it.

Managing swallowing problems

Models are starting to flag who needs a swallowing assessment: an Austrian hospital ran a dysphagia-risk model in its information system, but it missed more than half of at-risk geriatric patients, and the assessment itself stays with therapists.

One hospital's screening model; it does not show AI assessing swallowing or changing therapists' work.

Records, reports and coaching families

Inferred: in the large trial, volunteers and assistants supported practice and reported progress back to therapists, which shifts some coaching work; no record found measures how documentation is changing.

Inferred from one trial's design; no record measures documentation time or how coaching is done.