Dan Stephens, MA, LPC, LMHC, NCC, CSAT
A licensed counselor who builds and evaluates AI systems for clinical settings. Both sides of the table, which is the point.
I am a licensed counselor who builds and evaluates the AI systems used in clinical settings. I have sat on both sides of that table: carrying a caseload of twenty to twenty five sessions a week for years, and writing the rules and the tests that decide whether a model is allowed anywhere near a patient record. That is why I translate clinical nuance into product requirements without losing either side of it.
The things I have actually built are catalogued in the AI work portfolio, one page each, with what was built, what state it is in, and what it has not yet proven. The method behind them is written up separately.
Licenses and certifications
- Licensed Professional Counselor, Arizona, 2021 to present, with Clinical Supervisor authorization, 2024 to present
- Licensed Mental Health Counselor, Washington, 2019 to present
- Licensed Professional Counselor, Oregon, 2025 to present
- National Certified Counselor, 2016 to present
- Certified Sex Addiction Therapist, 2022 to present
Restored Life Counseling
Founder, Executive Director and Clinical Director. Glendale, Arizona, with clinicians in Arizona, Washington and Oregon. January 2018 to present.
I founded and lead a multi-state nonprofit counseling organization. I am board president and the clinical supervisor of record for five clinical contractors, including a psychiatric nurse practitioner. I carried a full caseload for years while building the organization’s clinical, governance and AI work. Revenue is up 24 percent year over year on a trailing twelve months. My clinical specialization is sex and pornography addiction, betrayal trauma, infidelity, and therapeutic disclosure.
Evaluating AI for clinical use
I built the checking gate that decides whether an AI-assisted clinical literature review is allowed out. I built it after giving the job to a commercial AI judge first and then auditing the judge. In its first real audit it flagged nineteen citations and was wrong nineteen times. Across a batch of thirteen reviews it raised roughly 359 objections, none of which survived checking, while passing every one of the five real errors as clean. The replacement is mechanical rather than a matter of judgment, it costs nothing to run, and it caught all five.
I wrote the hostile-review step that runs before anything outward-facing publishes: readers briefed to attack every dated or countable claim, each one checked against an original source, one recorded verdict apiece. The first real run examined 27 claims, caught 2 factual errors, and reworded 10 places where a claim had outrun its evidence.
Thirteen literature reviews and twenty six finished documents, about 165,000 words, have gone through those gates.
I wrote the organization’s policy on which AI may touch patient information, and by which route. The rule is that compliance is a property of a route rather than of a model: a specific model, served by a specific vendor, on a specific account, under a signed agreement covering that exact path. Consumer AI tools are prohibited outright. The same policy carries the read-provenance rule that what the AI read is never presented as what the clinician read.
I led a twelve-decision review of the crisis pathway in our screening instrument, including the decision to recommend against automated emergency dispatch.
Measurement and outcomes
I designed the Personal Challenges Survey and have fielded it since 2018: a self-report screener under fifteen minutes covering depression, anxiety, suicide risk, adverse childhood experiences, and problematic sexual behavior, with crisis-referral thresholds and a non-diagnostic design. I lead its validation and publication program, including scale reliability, factor analysis and prevalence norms, with a documented boundary that de-identifies all participant data.
Building the organization
Bylaws, provider manual, membership and business associate agreements, an institutional review board constitution and its operating procedures, and a records-retention policy. Monthly reporting to four audiences, and quarterly board governance as president.
Clineo LLC
Founder. Phoenix metro, Arizona. 2026 to present. A managed-services agency for small healthcare practices, built to handle patient information properly. Disclosed as an outside interest, with scope to be aligned to an employer’s interests at offer.
I designed its patient-data perimeter: every path closed by default, opened only when a signed agreement covering that exact path is verified and written down, with one-way flow and a person in the loop before anything identifying moves. No patient information enters the code at all. I also wrote the product specifications for its credentialing and claims lines, including a claims path whose sign-off checkpoints are bounded by an analysis of False Claims Act exposure.
Earlier
- Adjunct Instructor in Psychology, Warner Pacific University, January to May 2020
- Mental Health Counselor, Refresh Therapy, Vancouver, Washington, 2017
- Mental Health Counselor, Real Life Counseling, Vancouver, Washington, 2016 to 2017, and intern counselor there from 2014
Education
- MA, Counseling, Western Seminary, Portland, Oregon, 2016
- BS, Psychology, Washington State University Vancouver, 2012
- Graduate Certificate in Missional Theology, Missional Training Center, Tempe, Arizona, expected 2027
- Gottman Method Couples Therapy, Level II, The Gottman Institute
Also
I am designing a three-part continuing-education curriculum on AI ethics for counselors, with a provider application in development, including a reading of Arizona’s behavioral-health continuing-education rules.
The method writing behind the evaluation and governance work above is on this site, starting with the method page.
Getting in touch
Use the contact page. For counseling or clinical supervision, click here for a consultation.