If you’re a BCBA, BCaBA, or RBT, you’re already working with artificial intelligence (AI) in applied behavior analysis (ABA), whether your org has an “AI strategy” or not. The autocomplete in your email, the “smart summary” button in your telehealth platform, the dashboard that quietly flags a plateau in responding: all of these are AI systems touching ABA workflows.
That’s both exciting and unsettling. On a good day, AI can feel like an extra pair of hands helping with progress notes and goal drafting. On a bad day, it feels closer to a certain sci‑fi future where predictive systems try to decide what someone will do before they’ve done it. As behavior analysts, our mission is to use data to support meaningful behavior change, not to run a pre‑crime unit on our caseload.
This issue is about drawing a bright, practical line demarcating where AI is already embedded in ABA tools, which “jobs” it can safely help with, and which responsibilities must stay 100% human. The point isn’t to turn every clinician into a data scientist. It’s to give you language you can use with leadership, vendors, and each other when you’re asked, “Can we use AI for this?”
AI is already in your ABA workflow (even if you never bought an AI tool)
When people hear “AI in ABA,” they often picture a futuristic robot therapist. The reality is much closer to your daily to‑do list. The most common AI features BCBAs are encountering look like this:
1. Writing and goal-drafting helpers
-
“Rewrite this in parent‑friendly language.”
-
“Turn these bullet points into a paragraph for the treatment plan.”
These tools are advanced text prediction (see our blog on LLM-Wrappers for more on how these work). They’re great at smoothing grammar and expanding sketches into full sentences. They are not good at knowing what actually happened in session, which goals are appropriate, or what your funding source requires.
2. Note summarization and transcription
-
Live transcripts of parent trainings or team meetings
-
Draft SOAP‑style notes generated from your dictated bullet points
Here, AI is pattern matching over language samples, pulling out dates, decisions, and themes. Done well, this cuts the time from session to note. Done poorly, it can introduce subtle errors such as by changing “prompted once” to “prompted consistently,” or dropping context that matters for clinical judgment.
