PsychAssist | AI Report Writing in Psychological Assessment

Dr. Chris Barnes on what has to be true before you sign an AI-assisted psychological assessment report: the three types of AI tools in the field today, the six standards for defensible documentation, and a full live demonstration of a clinical reasoning engine working on a real configured workflow. Recorded 17 September 2026. — WHAT'S COVERED Every practice in assessment psychology carries a backlog, and that backlog clears at the speed cases close, not the speed they open. The bottleneck was never the expertise. It was the keyboard. Most tools in this field are solving for speed. Speed is the easy half. The half that matters is the half where you sign your name. This session covers the three categories of AI tool now being used in psychological and psychoeducational assessment — general-purpose AI, report generators, and clinical reasoning engines — what each one is genuinely good for, and where each one breaks. It then covers the six standards PsychAssist holds itself to for any AI-assisted assessment report: provenance, disconfirmation, clinician approval, clinical voice, auditability, and disclosure. The second half is a live demonstration on a configured ADHD assessment workflow: document extraction with clinician approval gates, supporting / convergent / divergent evidence surfaced against every diagnosis, cross-battery pattern detection, provenance on every claim, a defensibility check that reads the draft the way opposing counsel would, peer review and second opinion, the exportable audit trail, and automatic generation of secondary documents from the finished report. — 00:00 Welcome 03:44 Who I am, and what PsychAssist is 05:05 Every practice has a backlog — and why it's a documentation problem 06:16 Why speed is the easy part of this problem 06:46 What a clinical reasoning engine actually is 08:06 The three types of AI tools in assessment psychology 08:16 1 — General-purpose AI: ChatGPT, Claude, HIPAA-configured versions 09:09 2 — Report generators 09:43 Why divergence is the signal, not the noise 10:11 3 — Clinical reasoning engines 11:17 The faster typewriter that went to graduate school 11:46 The two things I need before I sign my name 12:19 Why we don't compete on a two-minute report 12:46 The six standards for defensible AI documentation 14:33 Into the platform 17:56 The case: Maya, 14, child ADHD assessment 19:16 Inside PsychAssist — workflows and the single source of truth 23:02 Document extraction and the clinician approval gate 24:49 Diagnoses: supporting, convergent and divergent evidence 27:27 Retaining clinical voice 28:39 Recommendation banks 29:08 Generating the report 30:13 NEW — Case conference with your whole file 30:25 The defensibility check: reading it as opposing counsel 31:05 Peer review 31:44 Second opinion on a diagnosis 32:58 Auditability, chat history and legal discovery 34:19 Styled report preview 35:43 Secondary documents from the finished report 36:54 Building a custom document live 39:18 Organisation-wide chat and practice metrics 40:23 What we covered 41:49 Q&A — What does "defensible" actually mean? 43:18 Q&A — Chat history, discovery and what got someone in trouble 44:53 Q&A — If something gets through, who is liable? 47:11 A report can't be assembled. It has to be derived. — ABOUT PsychAssist is an end-to-end clinical assessment platform for psychologists conducting psychological and psychoeducational evaluations — intake through signed report, and every document the report owes after it. Founded by Dr. Chris Barnes, a practicing clinical psychologist. https://psychassist.ai — A report can't be assembled. It has to be derived.