top of page

November 2026 Leading Edge Newsletter: Executive Foresight for Behavioral Health & Social Services Leaders

Writer: HiQuity Solutions
HiQuity Solutions
1 day ago
7 min read

The Newest Member of the System of Care: AI as a Social Agent in Behavioral Health Care


Introduction

Nearly one in five young people ages 12 to 21 now turns to an AI chatbot for mental health advice. That is about the same share who receive counseling from a mental health professional, according to a RAND study published in JAMA Pediatrics. Most of these young people, 63%, have told no one.

 

For most behavioral health organizations, the AI conversation still centers on operations: documentation, coding, scheduling, and the productivity gains that help a stretched workforce go further. Those questions matter. Almost no one is discussing what AI is doing to the people who use it.


Consider three people:


  • A young person consults a chatbot at midnight.

  • An adult talks to an AI companion more often than to friends.

  • A clinician revises a diagnosis after an AI pushes back.


Each of them is in a relationship with something that responds, remembers, and adapts.

 

AI Has Become a Social Agent

Researchers have known for three decades that people respond to computers socially. In The Media Equation, published in 1996, Stanford researchers Byron Reeves and Clifford Nass showed that people apply the same social rules to computers that they apply to one another, even when they know better.

 

Today's AI systems intensify that effect. They hold a conversation, recall what a person said last week, mirror tone, and offer encouragement on demand. A system that does those things functions as a social agent. When several of them work together, as agentic systems increasingly do, people find themselves inside synthetic social environments that someone else designed, for someone else's goals.

 

That shift changes the questions leaders need to ask. A tool is judged by whether it works. A social agent has to be judged by what it does to the people who interact with it: their trust, their judgment, their sense of agency, and their connections to other people. Those are the variables behavioral health understands best. They are also the variables most AI design leaves out.

 

The Evidence Is Arriving Faster Than the Field Is Adapting


Adults

A 2026 study in JMIR Mental Health surveyed 1,805 U.S. adults ages 18 to 49 and found that 35% used AI chatbots for mental health support at least one day in a typical week. People with moderate to severe depression or anxiety symptoms were more likely to use them. People reporting suicidal thoughts were more than twice as likely to be heavy users. Among AI users who had ever seen a mental health professional, 28% said they now see one less often. The authors are careful to note that a survey like this cannot show cause and effect.


Teens

A Common Sense Media national survey found that 72% of teens had used AI companions. About one in three teen users had discussed serious matters with an AI instead of a real person.


Dependence

In a four-week randomized study of 981 people by researchers at the MIT Media Lab and OpenAI, participants who chose to use the chatbot more showed more emotional dependence and more problematic use. Those who trusted the chatbot more, or felt more socially drawn to it, also showed more dependence.

 

None of these studies settles causation. Together, though, they describe a clear pattern. The heaviest use clusters among people who are already struggling, some of them are seeing human providers less, and many keep it to themselves. That is the population behavioral health organizations exist to serve.

 

A tool is judged by whether it works. A social agent has to be judged by what it does to the people who interact with it.

 

Clinicians Are in the Relationship Too

The social dynamics do not stop at the client. In a randomized trial posted as a preprint in 2025, 44 physicians completed 20 hours of AI training and then worked through clinical cases with an AI assistant. Physicians whose AI suggestions had been seeded with errors scored 14 points lower on diagnostic reasoning than those who received unaltered suggestions. The training did not protect them, and more experienced physicians declined more.

 

Deference to a confident voice is a well-studied social phenomenon. When that voice belongs to an AI system, it shapes clinical judgment the same way an authoritative colleague would, only faster and at greater scale. That makes it a supervision question before it is a technology question. Clinical supervisors are trained to notice how relationships shape judgment. Few have been asked to apply that skill to the AI relationships their staff now have.


The System of Care Has a New Member

System of care thinking starts from a simple premise: a person's well-being cannot be understood apart from the relationships and environments around them, including family, school, peers, and community. AI now belongs on that list.

 

Intake forms ask about substance use, sleep, and social supports. Few ask whether a client talks with an AI companion, how often, or what it tells them. The American Psychological Association's health advisory, issued in November 2025, urges clinicians to ask patients proactively about their chatbot use. It also asks them to watch for signs of unhealthy reliance, such as preferring the chatbot to people, hiding its use, or spending excessive time with it.

 

An organization that builds those questions into assessment, treatment planning, and care coordination will understand its clients more completely than one that treats AI as background noise.


Policy Has Started With Therapy Claims

States are moving quickly on one part of this picture. In 2026, Colorado, Maine, Rhode Island, Tennessee, and Vermont enacted laws restricting AI systems from delivering therapy or presenting themselves as mental health professionals, following Illinois and Nevada in 2025.

 

These laws address an important risk: products that claim to provide care they cannot provide. They do not reach the general-purpose chatbots that people use every day for advice and companionship. That gap is where behavioral health leaders have the most to contribute. The field that understands relationships, attachment, and judgment is best positioned to define what responsible AI looks like in people's lives, and to say so to policymakers, payers, and the technology companies building these systems.


What Forward-Looking Leaders Will Do

The shift from tool to social agent changes the questions leaders ask across the organization. The organizations that lead will make five moves:

 

  1. Assessment: They will add AI use to intake and treatment planning, so clinicians see the whole relationship landscape.


  2. Supervision: They will bring AI into clinical supervision as a relationship to examine.


  3. Vendor selection: They will judge vendors on the human variable as well as the technical one.


  4. Outcomes: They will measure what matters to people's lives, beyond what matters to throughput.


  5. Influence: They will use what they learn to influence state policy and payer expectations, rather than waiting for rules written by people who have never sat with a client in crisis.

 

Exhibit 1 shows how those moves change the questions leaders ask.

Table comparing a tool lens and a social agent lens on AI across five behavioral health leadership questions.

 

AI will keep getting more social, more persuasive, and more present in people's lives. Behavioral health can follow that change or help set its terms.

 

The field already has the theory, the clinical skill, and the standing to insist that the human variable be designed in from the start. The task now is to name what is happening, so we can prevent the harms and build the benefits on purpose.

 

Frequently Asked Questions


What does it mean to call AI a social agent?

It means recognizing that conversational AI systems respond, remember, and adapt in ways that trigger the same social responses people have toward one another. People extend trust to these systems, defer to them, and form attachments to them. Judging these systems only on accuracy misses their effects on judgment, agency, and human connection.


How many people use AI chatbots for mental health support?

Use is substantial and growing. RAND found that about one in five people ages 12 to 21 used chatbots for mental health advice in late 2025, up from about one in eight a year earlier. A 2026 survey of adults ages 18 to 49 found that about one in three used AI for mental health support at least one day a week.


Should clinicians ask clients about their AI use?

Yes. The American Psychological Association recommends that clinicians ask about chatbot use proactively and watch for signs of unhealthy reliance, such as preferring the chatbot to people or hiding its use. Building these questions into intake and treatment planning gives clinicians a fuller picture of the relationships shaping a client's well-being.


Which states restrict AI therapy?

In 2026, Colorado, Maine, Rhode Island, Tennessee, and Vermont enacted laws restricting AI from delivering therapy or presenting itself as a mental health professional, following Illinois and Nevada in 2025. These laws focus on therapy claims. They generally do not address the general-purpose chatbots people use for advice and companionship.


About the Author

Rhonda Bohs, Ph.D., is Systems Impact Strategist at HiQuity Solutions. She brings more than 30 years of behavioral health consulting experience in CCBHC compliance, System of Care frameworks, federal grant evaluation, and Medicaid-related behavioral health strategy, and has contributed to more than $500 million in funded federal grants. She also teaches social psychology.

 

HiQuity Solutions helps community-based organizations design, fund, and implement sustainable system transformation that enables them to do well and do good. Contact us at ask@hiquitysolutions.com, (612) 488-7337, or www.hiquitysolutions.com.

 

This newsletter is provided for general information and does not constitute legal, clinical, or professional advice. Research and policy details reflect sources as of October 29, 2026.


Sources



Comments

bottom of page