Every August, a new cohort walks into MSW, MS Counseling, MFT, PsyD, MD-psychiatry, PMHNP, and BCBA programs, and almost none will graduate with a single supervised hour of AI-assisted documentation practice. Then they show up at Jordan's group practice in Sacramento, or at an agency in Fresno like Carmen did, and they learn AI the way Carmen learned it: alone, at $32 an hour, paying $59 a month out of pocket for a scribe their supervisor did not know about. The pipeline produces clinicians for a workflow that no longer exists, and the gap gets filled by improvisation, which is how license-threatening habits form. This lesson teaches you, the experienced clinician with standing in the field, how to co-design the AI-native pre-licensure curriculum with the programs that feed your workforce: which competencies belong at graduation, how to map them to CACREP, COAMFTE, CSWE, and APA Commission on Accreditation standards, and how to anchor them in state pre-licensure hour rules at the BBS, OPP, BHEC, Florida DOH 491, and Illinois DPR. By the end you will have built a named artifact, the Pre-Licensure AI Competency Map, ready to bring to a program director's office.

The Flight School Problem: Training for the Cockpit That No Longer Exists

Hold one analogy through this lesson: a flight school that still teaches only paper charts and dead reckoning while every aircraft its graduates will fly has a glass cockpit. The school is not wrong that paper charts teach fundamentals; a pilot who cannot navigate without the screen is dangerous. But a school that never puts the student in front of the actual instruments is also dangerous, because the graduate's first encounter with the glass cockpit happens with passengers aboard. Aviation solved this decades ago: fundamentals first, then the instruments, then the failure modes of the instruments, and you are tested on all three before anyone hands you a license.

Pre-licensure behavioral health education is the flight school still teaching only paper charts. Students write practice notes by hand, learn diagnosis from casebooks, and role-play intakes, all of which builds fundamentals nothing should replace. But the workflow waiting at their first agency job includes AI scribes drafting notes from ambient capture, measurement-based care dashboards trending PHQ-9 and GAD-7 scores, and prior-authorization letters assembled with AI assistance. Their first encounter with these instruments happens with clients aboard, under a supervisor whose license is exposed by everything they do, often without any policy at all. Carmen's story is the normal case, not the exception: the associate adopts the tool first, the supervisor finds out later, and the supervision agreement signed eighteen months earlier says nothing about any of it.

The fix is not a webinar bolted onto the third year. The fix is curriculum co-design: practicing clinicians, the people who have survived payer audits and board inquiries, sitting with program faculty to define what an AI-competent graduate looks like, in the language accreditors and licensing boards already speak. That is the work of this lesson, and it is Level 5 work because it requires everything you built in Levels 1 through 4: you cannot teach a program what safe AI use looks like until you have practiced it, governed it, and defended it yourself.

Why Co-Design, Not Prescription

The first instinct of an experienced clinician looking at a training program is to hand the faculty a syllabus. Resist it. Graduate programs answer to accreditors, state licensing boards, universities, and students carrying significant debt, and they cannot add a course because a practitioner thinks they should. The AI competencies required at graduation get co-designed with MSW, MS Counseling, MFT, PsyD, MD-psychiatry, PMHNP, and BCBA training programs, not delivered to them. Co-design means you bring the field reality (the scribe stacks, the consent failures, the supervisor exposure, the payer documentation standards) and the program brings the pedagogical machinery (course sequences, practicum structures, accreditation self-studies, assessment rubrics).

Each degree pathway has its own constraints, and the competency map must respect them. MSW programs answer to CSWE and its Educational Policy and Accreditation Standards; counseling programs to CACREP; MFT programs to COAMFTE; doctoral psychology to the APA Commission on Accreditation; psychiatry residencies and PMHNP programs to their own medical and nursing accreditation structures; BCBA coursework to the BACB's verified course sequence requirements. An AI competency that cannot be expressed in the accreditor's existing competency language will never survive a curriculum committee, because programs document compliance against those standards in self-studies that take years to prepare. Your job is translation: take what the field needs and phrase it as an extension of competencies the accreditor already requires, such as ethical practice, documentation, supervision readiness, and technology-mediated service delivery.

There is a second reason co-design matters. Faculty are right to fear that AI instruction too early produces graduates who cannot write a note without the machine, the way a pilot trained only on autopilot cannot hand-fly an approach. The sequencing question (fundamentals first, instruments second, failure modes third) is pedagogical, and faculty are the experts in it. When you arrive prescribing, they hear a vendor. When you arrive co-designing, they hear a colleague who respects the part of the problem they own.

The Competency Architecture: What an AI-Competent Graduate Can Actually Do

A competency is a demonstrable capability, not a topic covered. "Students will be exposed to AI in behavioral health" is a topic; "the graduate can review an AI-drafted progress note against the session record and identify every unsupported claim before signing" is a competency. Build the map around capabilities a supervisor could verify in the first month of associateship, organized into five domains that recur across every degree pathway.

Domain one is documentation integrity: the graduate can produce a complete, medically necessary progress note without AI, then review an AI-drafted note word by word before signing, applying the cardinal rule this program has repeated since Level 1: the clinician signs the note, and the signature is a legal attestation, not a formatting step. Hand-flying first, instruments second. Domain two is consent and privacy: the graduate can explain to a client, in plain language, what an AI scribe does with session audio, knows that the HIPAA psychotherapy-notes carve-out and 42 CFR Part 2 create special categories that general AI tools were not built for, and can identify when a tool lacks a BAA. Domain three is risk boundaries: the graduate can state, without prompting, that AI never scores the CSSRS, never assigns a risk level, never makes the duty-to-protect determination, and never makes the mandated-report call; AI structures and formats after the clinician's determination, never before. Domain four is supervision transparency: the graduate discloses every AI tool in use to the supervisor before first use, because the supervisor's license is attached to the supervisee's documentation. Domain five is critical evaluation: the graduate can read a vendor claim and ask the questions that distinguish marketing from evidence, the skepticism this program describes as refusing any vendor claim that has not survived a board complaint or a payer audit.

Notice what is not in the architecture: no prompt-engineering tricks, no tool-specific training on Mentalyc or Upheal or Eleos as products. Tools churn; competencies persist. A flight school teaches instrument scanning, not one manufacturer's avionics menu. The map should name vendor categories so students recognize the landscape (generative scribes, measurement-based care platforms like Blueprint Health and Greenspace, EHR-native AI in SimplePractice and TherapyNotes) without certifying anyone on a product that may not exist when they graduate.

Train the fundamentals before the instruments, and train the failure modes of the instruments before the license. A graduate who can only document with AI is as unsafe as one who has never seen it.

Mapping to CACREP, COAMFTE, CSWE, and APA Accreditation Standards

Accreditation mapping is where most practitioner-led curriculum proposals die, so do it for the faculty rather than asking them to. The method is the same across accreditors even though the standards differ: find the existing standard that already obligates the program to teach something adjacent (ethics, documentation, technology in service delivery, supervision preparation), and express each AI competency as a specific instantiation of that standard. The program then gets to report your competency as evidence of compliance with a standard it already answers to, which converts your proposal from an added burden into an asset for the next self-study.

For a CSWE-accredited MSW program, documentation-integrity and consent competencies attach naturally to the ethical-practice and practice-evaluation competencies the EPAS framework requires every program to assess. For CACREP counseling programs, AI competencies map to the professional orientation and ethical practice foundations and to the technology-relevant expectations woven through the standards. For COAMFTE MFT programs, attach to the developmental competency components covering ethics, law, and technology-mediated therapy. For APA Commission on Accreditation doctoral programs, AI review skills slot into the profession-wide competencies in ethical and legal standards and in communication and interpersonal skills, where documentation already lives. For psychiatry and PMHNP training, the entry point is the documentation and systems-based practice expectations already embedded in medical education, plus the billing reality of 90833/90836/90838 add-on coding that residents must learn anyway. For BCBA coursework, the BACB's ethics requirements and the documentation demands of CPT 97151 through 97158, Medicaid review, and RBT supervision logs give AI competencies an immediate home.

In every case the deliverable is a two-column crosswalk: your competency on the left, the accreditor's standard number and language on the right, with one sentence explaining the fit. When the curriculum committee meets, that crosswalk is the difference between a polite decline and a pilot course.

Anchoring in State Pre-Licensure Hour Requirements

Accreditors govern the degree; licensing boards govern the hours, and the hours are where AI competency becomes enforceable rather than aspirational. Every pre-licensed associate (AMFT, APCC, ASW, post-doc) accrues thousands of supervised hours under board rules: the California BBS, the New York OPP, the Texas BHEC, the Florida DOH Board 491, the Illinois DPR. Those boards already require supervisors to oversee documentation and clinical judgment, and documentation now includes AI-assisted documentation whether the board has said the word or not. The curriculum map should therefore state, explicitly, which competencies must be demonstrated before the student begins accruing supervised hours, because the practicum and traineeship are where the student first touches real client records.

This anchoring has a practical payoff the program will appreciate: it answers the field-placement liability question. Agencies hosting practicum students increasingly run AI scribes, and a student who arrives with zero AI training is a liability the placement agreement never priced. A program that can hand its placement partners a list of verified AI competencies (this student knows the consent script, knows the risk boundaries, knows to disclose tools to the field instructor) makes its students more placeable, which is a currency program directors understand. Illinois deserves a special note in any map: the WOPR Act drew a statutory line around AI in therapy services, and Nevada's AB 406 did similar boundary work, so programs in those states have a legislated reason to teach where AI may and may not operate. Cite the statute for the state you are working in; never generalize one state's rule to another, because the boards do not.

The map should also flag the supervision handoff: the competency the student demonstrates in the program becomes the baseline the board-approved supervisor verifies in week one of associateship. That continuity, classroom to practicum to supervised hours to license, is the pipeline this chapter is named for, and the next lesson takes up the supervisor's side of it.

The Sequencing Spine: Fundamentals, Instruments, Failure Modes

Sequence is the part faculty will care about most, so propose one rather than leaving it implied. Year one (or the first phase of any program): fundamentals only. Students learn to write notes, conceptualize cases, and conduct risk assessment with no AI assistance, exactly as programs do now, and the map should say so in writing, because that sentence disarms the most common faculty objection before it is voiced. The pilot hand-flies first.

Phase two, typically alongside practicum preparation: instruments. Students work with AI drafting in a simulation environment, never on real client data, using standardized-patient transcripts or instructor-built cases. The teaching focus is the review discipline: take an AI-drafted note from a simulated session, find the three errors the instructor seeded into it (a fabricated quote, an unsupported clinical claim, a missing risk element), and correct them before "signing." This is the single highest-value exercise in the entire curriculum, because it builds the habit that protects every future client and every future supervisor: read every word before you sign, because the signature is the attestation.

Phase three, alongside or just before field placement: failure modes and governance. Students learn what happens when the tool is wrong, what a missing BAA means, why the free tier of a general chatbot is not a clinical tool, how 42 CFR Part 2 data demands stricter handling, what their disclosure obligation to a supervisor looks like, and how to read an agency AI policy (or recognize that the agency has none, which is its own red flag). Graduates leave able to do all three things: fly by hand, fly the instruments, and land safely when the instruments fail. That is the graduate Jordan wants to hire and the associate Carmen deserved to be.

Getting Into the Room: How a Practitioner Becomes a Co-Designer

Programs do not send invitations; you create the opening. Three doors reliably exist. First, the field-placement relationship: if your practice or agency hosts practicum students, you already have a contract and a contact, and field directors are the faculty members who feel the AI gap most acutely because placement sites keep asking them about it. Second, advisory boards: most accredited programs maintain community or professional advisory boards precisely because accreditors expect evidence of practitioner input into the curriculum; a seat on that board is the formal channel for the competency map. Third, adjunct teaching and guest lecturing: one well-built guest session on AI documentation ethics, taught with real artifacts (a redacted AI-drafted note with seeded errors, a consent script, a supervision-disclosure form), demonstrates the curriculum better than any memo and routinely converts into a standing module.

Walk in with the posture of the wise supervisor, not the disruptor. You are not telling faculty their training is obsolete; you are telling them the cockpit changed and offering to help them keep producing safe pilots. Bring evidence from your own practice: the verification checklist you actually use, the consent addendum that survived a payer audit, your own supervision-contract language. Practitioner artifacts carry an authority no slide deck matches, because they prove the competencies are practiced, not theorized. And accept the timeline: curriculum change moves in academic years, not quarters. A pilot module this year, a required unit next year, a self-study citation the year after: that is success, and it is worth the patience, because every cohort that graduates AI-competent is a cohort of supervisors-to-be who will never need to learn this the way Carmen did.

The Applied Problem: Build the Pre-Licensure AI Competency Map

Your artifact is the Pre-Licensure AI Competency Map: a single document, three to five pages, that you could hand to a program director at any of the seven degree pathways. Build it in four passes. Pass one: choose your target program type (start with the pathway you know best; an LCSW builds the MSW map, an LMFT the COAMFTE map) and write the five competency domains from this lesson as demonstrable capabilities, each phrased as "the graduate can..." with a verification method a field instructor could apply (direct observation, a seeded-error note review, an oral consent-script demonstration).

Pass two: build the accreditation crosswalk. Pull the relevant accreditor's published standards (CSWE EPAS, CACREP, COAMFTE, or APA CoA), and for each competency write the two-column row: your competency, the standard it instantiates, one sentence of fit. If you cannot find a plausible home for a competency in the existing standards, rewrite the competency until you can; that constraint is a feature, because it forces the field-need into language the program can act on. Use an AI assistant for the drafting if you like, with a prompt of this shape: "Here are five AI documentation competencies for pre-licensure training and here is the accreditor's standards language I have pasted in. Draft a crosswalk table mapping each competency to the most specific standard, one sentence of justification each. Do not invent standard numbers; use only the text provided." Then run the verification pass that everything in this program demands: check every standard citation against the source document yourself, because a hallucinated standard number in a curriculum proposal ends the conversation permanently.

Pass three: add the sequencing spine (fundamentals, instruments, failure modes) as a one-page phasing table tied to the program's existing structure: pre-practicum, practicum, field placement. State explicitly that phase one is AI-free. Pass four: anchor the state layer. Name your state board (BBS, OPP, BHEC, DOH 491, DPR), name the supervised-hour structure the graduates enter, and state which competencies must be verified before the student touches real client records, citing any state statute that draws an AI line, such as the Illinois WOPR Act or Nevada AB 406, only if it is your state's law.

Done looks like this: a program director who has never thought hard about AI can read the map in fifteen minutes, see that it costs no new faculty lines (it embeds in existing courses), see that it strengthens the next accreditation self-study, and see exactly which meeting to bring it to. That is a document that changes a pipeline.

Key Takeaways

• Pre-licensure programs are the flight school still teaching paper charts for a glass-cockpit workflow: graduates meet AI scribes, MBC dashboards, and AI-assisted payer documentation for the first time with clients aboard. The fix is curriculum co-design with MSW, MS Counseling, MFT, PsyD, MD-psychiatry, PMHNP, and BCBA programs, not a webinar bolted on at the end.
• Co-design beats prescription because programs answer to accreditors and boards, not to practitioners. Bring the field reality; let faculty bring the pedagogy. A proposal phrased in the accreditor's existing competency language survives the curriculum committee; a practitioner syllabus does not.
• Build five competency domains, each a demonstrable capability: documentation integrity (note-writing without AI, then word-by-word review before signing), consent and privacy (BAAs, the psychotherapy-notes carve-out, 42 CFR Part 2), risk boundaries (AI never scores the CSSRS, never assigns risk levels, never makes duty-to-protect or mandated-report calls), supervision transparency (disclose every tool before first use), and critical vendor evaluation.
• Map every competency to CACREP, COAMFTE, CSWE, or APA Commission on Accreditation standards in a two-column crosswalk, so the program can report your competencies as evidence of compliance with standards it already answers to. Verify every standard citation yourself; one hallucinated number kills the proposal.
• Anchor the map in state pre-licensure hour rules (BBS, OPP, BHEC, FL DOH 491, IL DPR) by stating which competencies must be verified before the student touches real client records, and cite state AI statutes like the IL WOPR Act or NV AB 406 only for the state they govern.
• Sequence as fundamentals, instruments, failure modes: phase one is AI-free skill building, phase two is simulated AI review with seeded errors, phase three is governance and failure handling. A graduate who can only document with AI is as unsafe as one who has never seen it.
• Enter through the doors that exist: field-placement relationships, program advisory boards, and guest teaching with real practice artifacts. Expect academic-year timelines, and measure success in cohorts: every class that graduates AI-competent never has to learn this alone at 9:54 PM the way the current workforce did.