AI Can Draft the Assessment Report. You Still Own the Evaluation.
Informational only; not clinical, legal or compliance advice. Check your payer policies, your clinic's approved-technology list and current vendor documents before you upload any client information.
The initial assessment report is where a lot of ABA authorizations are won or lost. It is also where clinics quietly default to a familiar hour band because the write-up is due and the waitlist is long. A newer set of tools will take the interviews, scores and notes you already collected and hand you a first draft. That can give you back an evening. It cannot choose the right assessment or decide what this child needs.
The draft is a write-up of work you already did. The evaluation, the recommended intensity and the medical-necessity story still have to be yours.
What these tools actually claim to do
Frontera Health's Assessment Builder, for example, states that clinicians upload intake interviews, assessment scores and notes, that its AI produces a first-draft initial assessment report, and that you can edit the report and export it as a Word file. The vendor lists support for several common tools, including the VB-MAPP, Vineland, ABLLS-R, AFLS, PEAK and SRS-2, and it describes a goal bank plus suggestions for goals and timelines. The listed price is $50 per assessment, with health-equity pricing the vendor says it offers for underserved or rural providers. Frontera Health, Inc. also publishes a Business Associate Agreement, last revised September 3, 2024. Read it before you upload records.
The vendor's FAQ is clear on the point that matters: its models create a first draft, and "you, the professional, should review our reports." The same FAQ says the tool can summarize video or audio of parent or caregiver interviews. If you record a family, settle consent first; we covered that conversation in an earlier post. Do not treat marketing claims about time saved or approval rates as evidence.
What you still own
The BACB Ethics Code for Behavior Analysts puts the assessment itself, not just the write-up, on the behavior analyst. Standard 2.13 says you select, design and implement assessments that are conceptually consistent with behavioral principles, based on scientific evidence and matched to the client's and stakeholders' needs, context and resources, then summarize the procedures and results in writing. Standard 2.14 says the interventions that follow have to be based on those results.
The Council of Autism Service Providers' Practice Parameters for Artificial Intelligence Use in Applied Behavior Analysis make the same split. The practitioner is responsible for clinical documentation, including full clinical responsibility to prevent harm and to document effective, individualized recommendations. In CASP's words, any AI-powered work product should align with the principles of behavior analysis and reflect individualized treatment decisions, and AI should support clinical work, not supplant it.
That frame matters even more at intake. In August 2026, CMS released a state toolkit on ABA services and said the point is care that is medically necessary and individualized, with treatment intensity that reflects each child's clinical needs rather than a standardized service model.
A BCBA review before anything goes out
- Scores and procedures are what you administered. If the draft cites a protocol, subdomain or age equivalent you did not collect, take it out. CASP warns that generative tools can confabulate, filling a gap with a plausible but wrong target, or invent details that were never in the record.
- Interview summaries match what the family said. Read any AI summary against the recording or your notes. Correct anything that overstates or understates a caregiver concern.
- Goals are this child's goals. A goal bank is a menu, not a plan. Replace generic skill lists with targets that matter where this person actually lives.
- Recommended intensity is a clinical argument, not a default band. If the draft proposes hours or a timeline, ask whether those hours follow from the findings, the barriers and the goals.
- The medical-necessity chain is visible. Assessment findings, prioritized goals, proposed intensity and how you will know it is working should read as one story.
- No PHI outside approved tools. Do not paste evaluation records into a general chatbot to "clean up" a report.
- You still sign, and you can say a tool helped. CASP expects clinicians to disclose AI use in a work product where it contributed to the content. The BCBA who completed the evaluation owns the report.
The useful trade is the same one as with treatment-plan drafts: let the tool assemble the shell, then spend the time you saved on the part no model should own. Does this report describe the child you assessed, and is the requested intensity justified for this person, now?
Let the tool draft. You evaluate.
Sources
- https://fronterahealth.com/assessment-builder/
- https://fronterahealth.com/baa
- https://www.casproviders.org/practice-parameters-for-ai
- https://assets-002.noviams.com/novi-file-uploads/casp/pdfs-and-documents/AI_Guidelines/aipracticeparameters.pdf
- https://www.bacb.com/wp-content/uploads/2022/01/Ethics-Code-for-Behavior-Analysts-240830-a.pdf
- https://www.cms.gov/newsroom/press-releases/cms-launches-new-state-toolkit-protect-children-autism-strengthen-oversight-applied-behavior
More from the blog
- When a Payer Uses AI on Your ABA Authorization: What New State Laws Require
- CASP Demo Days Are Next Week. Here Are the Questions Worth Asking.
- SOC 2, BAAs and Badges: How to Read an AI Vendor's Security Claims Before Your ABA Clinic Signs
- An Autism Center Replaced Its Supervisor With AI. Here Is What AI Can't Take Over in RBT Supervision
Informational only; not legal, clinical or compliance advice. Vendor HIPAA/BAA statements are as published by the vendors and are not verified by us. Sponsorship never changes our rankings. Corrections: admin@gtbehaviorbolts.com.