I didn’t experience BHT the way I expected to.
A few medical issues on our team meant I spent most of the conference flying solo at our booth. That included learning, somewhat last minute, how to set the thing up, how to tear it back down, and everything in between.
It was outside my comfort zone.
It also turned out to be one of the most valuable (and exhausting) parts of the trip.
Because I couldn’t disappear into sessions for most of the day, I spent a lot of time talking to people.
I talked with health plan CEOs and healthcare executives. I talked with clinicians and people building behavioral health companies. I talked with college students hoping to become psychologists.
Different backgrounds. Very different vantage points on the industry.
But one topic came up constantly: AI.
That part wasn’t surprising. AI was everywhere at BHT.
What struck me was how often the conversations eventually arrived at the same question:
How can we use AI to make behavioral healthcare more human, rather than getting in the way of care?
There is understandably enormous interest in whether AI can provide behavioral health support directly.
And I think some of those tools will be genuinely valuable.
For lower-severity needs, an AI system can be a useful sounding board. It can help someone work through their thoughts, learn basic techniques, prepare for a difficult conversation, organize what they want to discuss with a clinician, or simply take a first step toward addressing something they otherwise might have ignored.
In a behavioral health system with persistent access challenges, we shouldn’t dismiss that value.
But I’m increasingly convinced that the most important AI opportunity in behavioral health isn’t replacing the therapist at all. It’s using AI to strengthen the people, relationships, and systems already responsible for delivering care.
The opportunity is to make those human connections more informed, more effective, and easier to access.
Giving providers more leverage
Clinicians spend an extraordinary amount of their time doing work around care rather than providing care.
They gather context. Review records. Document sessions. Search for resources. Follow up with patients. Keep up with evolving evidence and clinical practices.
You could see where the market thinks AI fits just by walking the BHT expo floor.
There were EHR booths everywhere, and I’m not sure I saw a single one that didn’t have “AI” somewhere on the backdrop.
That makes sense.
Documentation and administrative burden are some of the most obvious places where this technology can create immediate value.
AI can increasingly shoulder some of that work.
Imagine a provider walking into a session already understanding what has changed since the last visit, what assessments or patient-reported outcomes suggest, what topics may deserve attention, and what evidence-based approaches might be relevant.
Or leaving the session without another pile of administrative work waiting for them.
But I think the opportunity is bigger than simply making the EHR faster.
The point isn’t to automate the relationship.
It’s to give the clinician more capacity to participate in it.
That’s also a big part of how we think about AI at Psych Hub.
One of the areas I’m most excited about is using AI to help providers build skills before they’re sitting across from a client.
Our therapy simulator, for example, gives providers a chance to practice real clinical scenarios, get reps in, and receive feedback in a low-risk environment.
That’s a very different application of AI than trying to insert a model into the therapeutic relationship itself.
The technology creates space for the provider to practice, learn, make mistakes, and improve before the human interaction happens.
To me, that’s one of the clearest examples of AI empowering better care without adding friction to the relationship.
More broadly, we should be thinking about AI as a way to continuously upskill and equip providers – putting evidence, education, feedback, and decision support into their hands at the moments they actually need it.
The opportunity isn’t just to save clinicians time.
It’s to help them become better clinicians.
Patients are going to use AI whether we prescribe it or not
There’s another reality the industry has to confront.
Patients are already talking to these systems.
That means a provider may increasingly need to understand not just what medications someone is taking or what other clinicians they are seeing, but what AI tools they are using and how they are using them.
What are you asking the system?
What kinds of advice are you taking from it?
What role is it playing in the way you understand your own mental health?
That’s not necessarily a bad thing.
But we’ve also already seen examples in which conversations with general-purpose AI systems have reinforced harmful beliefs or delusions rather than appropriately challenging them or directing someone toward care.
So part of the provider’s role may eventually become helping patients understand the difference between productive and inappropriate uses of these tools.
In other words, AI literacy could become part of behavioral healthcare itself.
That also reinforces why the answer can’t simply be “AI is good” or “AI is dangerous.”
The more useful question is when a particular tool is appropriate, for which person, at what level of severity, and when the next step should be a higher level of care.
The behavioral health system is bigger than behavioral health providers
One of the other things that became clearer to me throughout the week is how limiting it can be to think about behavioral healthcare as something that happens exclusively between a patient and a therapist.
That’s not how many people enter the system.
They talk to a primary care doctor.
A teacher or school counselor.
A faith leader.
A coach.
A family member.
Someone they trust within their community.
Those human connections are already part of the behavioral health ecosystem, whether or not we formally label them that way.
If we’re serious about improving outcomes, we should think about how technology can better equip those people too: helping them recognize when someone may need support, understand what resources are appropriate, and know when the situation requires a higher level of care.
And that support can’t assume every community interacts with behavioral health in the same way.
Cultural competency, trust, language, and community norms aren’t edge cases in behavioral health.
They’re part of the care environment.
The opportunity isn’t just to build better tools for clinicians.
It’s to strengthen the network of people surrounding someone before, during, and after formal treatment.
Innovation needs pathways
That brings me back to one of the few sessions I was actually able to participate in at BHT.
Our panel discussed an HHS pledge bringing organizations across the industry together around behavioral health clinical care pathways.
At first glance, clinical standards and rapid AI innovation might seem like two separate conversations.
I think they’re increasingly the same conversation.
If we’re going to introduce new AI tools across behavioral healthcare, we need a clearer shared understanding of what appropriate care looks like.
When is self-guided support appropriate?
When should technology encourage someone to involve another person?
When does someone need a licensed clinician?
When should care escalate because severity or risk has changed?
And once someone enters treatment, what should high-quality, evidence-based care actually look like?
Standards aren’t there to prevent innovation.
Done well, they give us the guardrails to innovate faster and more responsibly.
They can help us make better decisions about where AI belongs, where it doesn’t, and when a person needs to move from one level of support to another.
A disruptive moment
Behavioral health feels like it’s in a genuinely disruptive period.
You could feel it at BHT.
Everyone is moving quickly.
Companies are experimenting. Providers are figuring out where these tools belong. Health plans are evaluating entirely new categories of solutions. EHR vendors are racing to integrate AI. And the boundaries between education, self-care, coaching, clinical treatment, and healthcare navigation are becoming less obvious.
There will be mistakes.
Some products will overreach. Some applications of AI won’t work. We’ll discover risks we haven’t anticipated yet.
But I left the conference optimistic.
Because the future doesn’t have to be a choice between technology and human connection.
With thoughtful clinical standards and the right guardrails, AI can help patients get support earlier.
It can help the people around them recognize when more help is needed.
It can help clinicians practice, build skills, and continuously improve before they ever walk into the room with a client.
It can give providers better context before a session and reduce the administrative burden waiting for them after one.
And, ideally, it can give clinicians more time and attention for the part of behavioral healthcare that technology is supposed to support in the first place: the relationship with the person sitting across from them.
We’ve spent a lot of time asking whether AI can provide behavioral healthcare.
The question I’m leaving BHT more interested in is:
What becomes possible when we use AI to make every human connection around a patient better?
Leave a Reply