Two years ago, AI features didn't exist on most ABA software platforms. Now, it's hard to find one that doesn't mention AI somewhere on its homepage.
A lot of what AI is aiming to do is genuinely useful, but it’s a large change for the industry to absorb in such a short time. It also raises a question that’s absent in most feature comparisons: What determines whether an AI feature actually saves your team time, versus giving them one more draft to review and rewrite from scratch?
The language model choice matters, but the bigger factor is something most feature pages don't talk about: What data the AI has access to. An AI feature is only as good as the data it can act on. When it sees more of your practice, it has potential (when built right) to produce something closer to what your team would do themselves. When it has less data than is needed, it fills in the gaps with generic language, poor or unsafe assumptions, and output that needs heavy editing before anyone would dare stake their clinical reputation on it.
What separates useful AI from generic outputs
Throughout the industry, ABA practices are adopting AI tools at a pace we haven’t seen with any previous technology. Flychain's 2026 AI Adoption & Spend Report found that the share of practices paying for AI tools went from 19% to 50% in about twelve months. All the while, the platforms those practices run on are shipping their own AI features: note drafting, scheduling, billing, and clinical coaching.
Practices want AI to reduce the admin load, and the ABA AI market is racing to deliver it.
But there are myriad examples of what gets missed in the race to deliver. Consider an AI scheduling feature that can only see staff calendars, it will propose times that are technically “open.” Give that same AI feature access to authorization utilization, client treatment-related requirements, and provider credentialing status, and it begins to suggest times that are actually schedulable. An AI feature built like this is much less likely to create a problem for your organization down the road because someone's authorization was running low or a credential had expired.
A billing agent that only has claims data may be able to submit those claims. Yet one that also has access to the session data behind each claim, the authorization tied to each appointment, and the credential standing of the rendering provider can flag issues before submission. That's the difference between an AI that catches a preventable denial before submission versus you being surprised to find out in a report sixty days later.
A session note assistant that knows a session lasted ninety minutes and a CPT code was billed writes a note that could apply to almost anyone. One that knows which programs were run, what targets were addressed, what data were collected, and what prompting levels were used writes a draft that sounds like it came from someone who observed the session.
The principle is the same across every workflow. When AI has access to more of the context your team already works with, it produces output that's closer to done. When it doesn't, your team ends up spending time fixing what the AI got wrong, or filling in what it couldn't know, and the time savings start to shrink.
The difference between connected data and siloed data
It may be surprising, but how your ABA platform is built matters more than any singular feature.
When scheduling, data collection, billing, credentialing, and authorization management live in separate systems or modules that don’t talk to each other, every AI feature works from a partial picture. It may do an okay job within its own silo, but it can't understand relationships across workflows, and errors compound.
Instead, when those systems are able to share the same data layer, the AI has what it needs to follow the thread. Learner data connects to a session. A session connects to a schedule. Scheduled hours connect to an authorization. The authorization connects to a credential. The credential connects to a claim. Each of those connections carries information that makes every AI feature in the chain more accurate and more useful.
In practice, this is the difference between an AI that flags authorization utilization running behind after sessions are already on the calendar versus one that catches the trend before the next appointment is booked.
The question worth asking about any AI feature in ABA software is not if it exists, but "How much of my practice does this feature actually understand when it runs? What data is it able to safely access?"
Richer data doesn't mean exposed data
That brings us to the next point…If AI works better when it has access to more of your practice data, the natural next concern is: what happens behind the scenes with that data?
In ABA, we're talking about clinical information involving children and families. Financial data tied to insurance and billing, and operational details about staff, schedules, and authorizations.
The distinction we’re making here is that using connected practice data to produce useful output for the people who already have access to that data is not the same as exposing it to something outside your practice. The AI should work within the same environment your team already uses, within the protections it should already have in place. The data shouldn’t leave to go somewhere else so the AI can “think” about it.
How Motivity Intelligence follows this principle
This is the reasoning behind Motivity Intelligence, the AI layer across our platform. Because clinical data collection, scheduling, billing, credentialing, and authorization management live in the same platform, each AI feature can pull from the context it needs when it runs. An authorization tool can reference scheduling data. A billing feature can check session records.
Each feature is more useful because it can reach the data that's relevant to its job, without anyone having to export, reconcile, or copy information between systems.
Jenny Saavedra, owner at ABA Spectrum Therapy, described what that feels like from the practice side:
"With the AI coach in there, I can quickly glance and see how my BCBAs are doing on utilization. That's the piece I love. It's transparent there. I don't have to pull a report to tell a BCBA, 'Here's what you've utilized.' They can do that themselves and they can tell me how they're doing on their utilization, which directly impacts that client progress. And they can see that too."
That's the version of AI in ABA we think is worth building. Not the one with the longest feature list, but the one where each AI feature has been thoughtfully designed with the context it needs to do the best work.
When you evaluate AI in ABA software, look at the foundation first
The most useful AI in ABA won't come from whoever ships the most features. It'll come from whoever gives those features the most complete picture of what's happening across a practice.
When your team looks at AI capabilities in ABA software, the features themselves will get a lot of attention. That's fair. But before comparing feature to feature, look at what's underneath: how deeply does this platform understand what happens across your sessions, your schedules, your authorizations, and your billing? Can the AI see across those workflows, or is each feature working on its own?
That foundation is what determines whether AI saves your team real time or just moves the editing work from one screen to another.
If you want to go deeper, I’ll be talking through what we've launched, what we decided not to build, and how to tell a real AI capability from a feature name in an upcoming webinar.
Register for the webinar: AI in ABA, September 8th →
