Building AI for Therapy Practices

Behind the Scenes with Uriah Guilford

I’ve spent the last year talking with therapists about artificial intelligence. Most of those conversations tend to revolve around the same questions.

Is it HIPAA compliant? Is it secure? Is this something I should even be using? What happens if it gets something wrong? And perhaps most commonly, does AI belong in a therapy practice in the first place?

They’re good questions. They’re also questions I expect we’ll continue asking for quite some time.

Recently, though, I realized there was a perspective I hadn’t spent much time exploring.

What does it actually look like to build AI specifically for therapy practices?

There were certainly things I was curious about from a technology perspective. I wanted to understand how crisis situations were handled, what kinds of safeguards had been put in place, and what surprised someone after spending the last year building and refining a platform used by therapists. But I was also interested in something else: what does the process of building AI in a healthcare environment actually teach us?

That question led me to a conversation with therapist and entrepreneur Uriah Guilford.

Many therapists know Uriah through Productive Therapist, where he has spent years helping practice owners improve intake processes and practice operations. More recently, he has been focused on developing an AI-powered intake and receptionist platform for therapy practices, an experience that has given him a front row seat to many of the same questions therapists are currently wrestling with.

As we talked, I was struck by how little of our conversation centered on AI being “the future” of therapy. Instead, we found ourselves talking about things that will probably sound familiar to most therapists: implementation challenges, privacy considerations, human oversight, and the reality that even useful technology has limitations.

In some ways, I found that reassuring. There is no shortage of headlines about AI. Depending on the day, we’re either being told it’s going to revolutionize healthcare or replace entire professions. What I heard during our conversation felt much more grounded. Building AI for therapy practices involves making decisions about boundaries, workflows, risk, and responsibility. In other words, many of the same things therapists think about every day.

And maybe that’s part of the reason I found the conversation so interesting.

The longer I spend talking with therapists about AI, the more I find myself coming back to questions of judgment and discernment. Not because there’s always a single right answer, but because thoughtful implementation almost always involves weighing competing priorities. Privacy and convenience. Innovation and caution. Efficiency and human connection.

Those tensions showed up repeatedly throughout our conversation, and I suspect they’ll continue showing up as therapists decide what role AI should play in their practices in the years ahead.

Why Build AI for Therapy Practices in the First Place?

Uriah didn’t set out to build an AI platform.

Like many therapists who eventually find themselves wearing multiple hats, he started by trying to solve a problem.

After beginning his career as a therapist and later growing a group practice, he found himself facing a challenge that will sound familiar to many practice owners: intake. More specifically, how do you respond to prospective clients quickly when people are often reaching out at the exact moment they’re ready to seek help?

That question ultimately led to Productive Therapist, his virtual assistant company, which has spent the better part of the last decade supporting therapists with intake coordination and administrative workflows. Over time, another question kept surfacing: Could someone answer the phone live?

It was a request he heard repeatedly from therapists. Practices wanted faster response times and more immediate support for prospective clients, but providing live coverage across multiple practices isn’t always easy to scale with a human team alone.

Then, about a year ago, he began experimenting with AI voice agents.

“I was surprised that they had gotten so good,” he told me.

That observation eventually led to the development of an AI-supported intake platform that is now being used by dozens of therapy practices. But what interested me wasn’t necessarily the platform itself. It was the path that led there.

At its core, this wasn’t really a story about artificial intelligence. It was a story about trying to solve a problem that many therapists have experienced themselves.

I keep coming back to that because I think it’s easy to get distracted by the technology. There is so much happening, so quickly, that it can sometimes feel like we’re all trying to catch up. Every week seems to bring another tool, another announcement, or another promise about how AI is going to save us time and transform the way we work.

Toward the end of our conversation, I asked Uriah what question he wishes every therapist would ask before implementing an AI tool.

His answer was immediate:

“What are you trying to accomplish?”

It’s a simple question, but I suspect it’s one many of us don’t spend enough time with. What problem are we trying to solve? What would success actually look like? And does AI meaningfully help us get there?

Ultimately, I don’t think most therapists are looking for more technology. I think they’re looking for ways to better support their clients, protect their practices, and make thoughtful decisions about where technology fits into the work they do every day.

What Therapists Don’t See Behind the Scenes of AI

One of the things that stayed with me after our conversation was how often Uriah returned to the topic of testing.

I suppose I expected to hear more about capabilities and features. Instead, we spent a surprising amount of time talking about edge cases, refining prompts, and asking questions like, “What happens if someone says they’re suicidal?” or “What happens if the AI is asked something completely unexpected?”

By the time most therapists encounter an AI tool, we’re usually seeing something that appears relatively polished. We aren’t seeing the months spent trying to anticipate problems before they happen, the conversations about limitations, or the decisions that go into determining what a system should and shouldn’t do.

“We’ve tested it extensively and tried to break it,” Uriah told me. “What happens if somebody says this? What happens if they say they’re suicidal? What happens if they ask this random question?”

I found myself thinking that this probably isn’t something most therapists consider when they’re evaluating AI tools, but perhaps it should be.

We’re often taught to ask whether a platform is HIPAA compliant, whether a Business Associate Agreement is available, and how Protected Health Information is handled. Those questions matter. But listening to someone describe the process of building these systems made me wonder if there are additional questions worth asking.

How was it tested? What assumptions were made during development? What happens when the tool encounters something it wasn’t designed to handle?

Those questions won’t necessarily tell us whether a particular platform is right for your practice, but they do provide insight into how thoughtfully it has been developed.

Another thing that struck me was how intentionally limited the platform was designed to be.

That may sound counterintuitive at a time when so many AI tools are being marketed as capable of doing almost anything, but Uriah repeatedly emphasized the importance of keeping the AI focused on a relatively narrow set of tasks. Scheduling appointments and answering intake questions is one thing. Exercising clinical judgment is something else entirely.

As therapists, we’re already accustomed to thinking about scope. We understand that being highly skilled in one area doesn’t automatically translate into competence in another. In some ways, I found myself thinking about AI similarly throughout our conversation. What is this tool designed to do? Just as importantly, what isn’t it designed to do?

I suspect those questions will become increasingly important over the next few years.

There is a tendency to talk about AI as though it’s one thing, but it isn’t. A voice agent handling appointment scheduling presents a different set of considerations than an AI documentation assistant, which is different from a chatbot, which is different from a clinical decision support tool.

Those distinctions matter.

And perhaps that’s one of the things I appreciated most about our conversation. There was very little discussion about building an AI that could do everything. Instead, we spent a lot of time talking about limitations, boundaries, and where human judgment still belongs.

As therapists, I suspect we’re going to spend a lot of time having those same conversations ourselves.

One Thing That Stayed With Me

Throughout our conversation, I was struck by how often discussions about AI became discussions about boundaries. What should this tool do? What shouldn’t it do? Where does responsibility shift back to a human being?

Those aren’t just questions for people building AI. They’re questions therapists will likely continue asking as these tools become more common in practice.

Can AI Handle Crisis Situations?

At some point, every conversation about AI in mental health seems to arrive at the same question: What happens in a crisis?

It’s a fair question, and one that therapists tend to ask almost immediately. Whether we’re talking about documentation tools, chatbots, or AI receptionists, there’s an understandable concern that eventually someone will disclose something urgent or unexpected.

That was one of the first things I wanted to understand.

According to Uriah, crisis scenarios were considered early in the development process. If someone indicates they are suicidal or experiencing an emergency, the platform is designed to provide pre-determined responses directing individuals to appropriate crisis resources and emergency services.

What I appreciated about his answer was that it wasn’t presented as a technological triumph. It was presented as a boundary, and I think that’s an important distinction.

Throughout our conversation, I found myself noticing how often discussions about AI eventually became discussions about limits. Not in a negative sense, but in a practical one. What should this tool do? What shouldn’t it do? Where does responsibility shift back to a human being?

Those questions aren’t unique to AI. Therapists ask some version of them all the time.

When is referral appropriate? What falls outside my scope of practice? When do I consult? At what point do I need additional support?

In many ways, implementing AI responsibly requires a similar kind of thinking.

An AI receptionist can help answer questions, provide scheduling information, and assist with intake processes. It cannot assess risk, provide therapeutic support, or exercise clinical judgment, nor should it. That doesn’t necessarily mean AI has no place in a therapy practice. It simply means we need to be thoughtful about where we expect it to operate.

I also think it’s worth acknowledging that crisis situations aren’t unique to AI. Human receptionists can miss information. Emails can go unread. Voicemails can sit unheard over a holiday weekend. None of us are operating in environments completely free from risk.

The question, at least in my mind, isn’t whether AI can eliminate risk entirely. It’s whether we understand the risks well enough to implement these tools thoughtfully and establish appropriate boundaries around their use.

That’s one of the reasons I continue to come back to discernment.

There will likely never be a checklist that answers every question about AI in healthcare. Context matters. Practice settings differ. Therapists have varying levels of comfort with technology. What makes sense for one practice may not make sense for another.

But I do think there are certain questions worth asking regardless of the tool:

What is this technology designed to do?
What happens when something falls outside that scope?
Where does human oversight begin?
How are exceptions handled?
Are those answers consistent with my values and responsibilities as a therapist?

Those aren’t necessarily easy questions. Then again, some of the most important questions in our profession rarely are.

And if my conversation with Uriah reinforced anything, it’s that the people building these tools are asking many of the same questions therapists are asking when deciding whether to use them.

Why Human Oversight Still Matters

If there was one theme that surfaced repeatedly throughout our conversation, it was this: AI still requires people.

That may seem obvious, but I think it’s easy to lose sight of that when conversations about AI tend to alternate between two extremes. Depending on who you ask, AI is either going to replace large portions of the workforce or completely transform the way we work.

As is often the case, the reality appears to be much more nuanced.

At one point, Uriah compared AI to a pre-licensed therapist. The analogy made me laugh at the time, but I understood what he was trying to convey. The technology may have significant capabilities, but it still requires guidance, oversight, and ongoing refinement to function effectively.

That was particularly evident when we started talking about implementation. According to Uriah, one of the biggest challenges wasn’t getting the technology to work. It was helping therapists and intake coordinators learn how to work with it. Users needed to understand how to monitor conversations, identify situations that required adjustments, and communicate those refinements back to his team.

In other words, implementing AI still involved people.

I found that interesting because I think many of us have been conditioned to think about AI as a way to remove human involvement. Yet much of what he described sounded less like replacement and more like supervision, and that distinction feels important.

Therapists are already accustomed to supervising and mentoring. We understand that competence develops over time and that even experienced professionals benefit from consultation and feedback. Listening to Uriah describe the implementation process, I found myself thinking that AI adoption may require a similar mindset. Not because AI is a person, of course, but because introducing any new system into a practice requires oversight.

Someone still needs to ask:

Is this working as intended?
Are there patterns we’re missing?
Does anything need to be adjusted?
Is this improving the client experience?
Are we becoming too reliant on the technology?

Those aren’t really AI questions. They’re implementation questions, and they don’t disappear simply because a platform is functioning as designed.

I was also struck by something else during our conversation. At no point did Uriah suggest that therapists should stop being therapists. If anything, his perspective seemed to reinforce the opposite. The more we talked about AI, the more it became apparent that many of the responsibilities therapists already carry, exercising judgment, establishing boundaries, and protecting client welfare, remain firmly in place.

Technology may change. Professional responsibility doesn’t.

I suspect that’s one of the reasons conversations about AI can sometimes feel uncomfortable. There is often an underlying hope that a tool will make difficult decisions easier or remove some of the uncertainty that naturally comes with running a practice.

In my experience, that hasn’t been the case.

If anything, AI seems to ask us to become even more intentional and more willing to ask difficult questions. Not because therapists need to become AI experts, but because we’ll likely need to become increasingly comfortable making thoughtful decisions about technology in environments where the answers aren’t always obvious.

That, more than anything else, was probably my biggest takeaway from our conversation.

Human oversight isn’t disappearing. If anything, it’s becoming more important.

What Are the Limits of AI in Therapy Practices?

One of the things I appreciate about talking with people who build technology is that they tend to develop a healthy respect for its limitations, and that certainly seemed to be true during my conversation with Uriah.

At various points, we talked about Electronic Health Record integrations, privacy concerns, and the challenges that come with moving Protected Health Information between systems. Interestingly, some of the hesitation didn’t come from a lack of technical capability. It came from asking whether certain things should be done in the first place.

For example, Uriah shared that his team has been intentionally cautious about direct EHR integrations. While therapists often appreciate the convenience of having systems communicate with one another, every additional connection creates another consideration from both a privacy and security perspective.

I found that particularly interesting because I think many of us tend to assume that more integration is automatically better. Sometimes that’s true. Sometimes it simply introduces additional complexity, and the same could probably be said for HIPAA.

At one point during our conversation, Uriah joked that, as someone who enjoys technology and innovation, there are moments when he imagines how much easier life would be if he were building tools for roofers instead of therapists. I laughed because I suspect many people who work in healthcare have had a similar thought at one point or another.

To be clear, he wasn’t arguing against HIPAA. In fact, he was quick to acknowledge why those protections exist. But his comments highlighted something I think is worth acknowledging: innovation and regulation don’t always move at the same pace.

Therapists experience this tension all the time. A new tool is released, colleagues begin talking about it online, and suddenly we’re trying to determine whether it’s something we should even consider bringing into our practices. Is it secure? Will the company sign a Business Associate Agreement? How is data stored? What administrative safeguards are needed?

The reality is that healthcare simply operates differently than many other industries, and for good reason.

One of the things I appreciated most about our conversation was that it never felt like an argument against those responsibilities. If anything, it reinforced them. Building technology for healthcare means accepting that privacy, security, and risk management are part of the job. They’re not obstacles to work around. They’re part of the work itself.

I don’t know that this is something we’ll ever solve once and for all. If anything, I suspect it’s an ongoing process of asking thoughtful questions and making decisions based on the information available to us at the time.

And perhaps that’s another lesson worth taking from this conversation.

The people building technology for therapists aren’t exempt from those questions. In many cases, they’re asking the same ones.

Stay Updated on Compliance Changes

Compliance expectations are constantly evolving, and most providers don’t hear about changes until they become a problem.

If you want clear, practical updates you can actually use, you can join my email list below.

What Question Should Therapists Ask Before Using AI?

Toward the end of our conversation, I asked Uriah if there was one question he wished every therapist would ask before implementing an AI tool.

His answer was immediate:

“What are you trying to accomplish?”

I’ve found myself thinking about that question quite a bit since our conversation.

It’s deceptively simple, but I suspect it gets at the heart of many of the conversations therapists are currently having about AI. We’re often trying to determine whether a tool is good or bad, helpful or harmful, appropriate or inappropriate. Those questions matter, but they can sometimes pull us away from a more fundamental one.

What problem are we actually trying to solve?

For one therapist, the answer might be improving response times for prospective clients. For another, it may be reducing documentation burden or streamlining administrative tasks. Some therapists may decide that AI has a meaningful place in their practice, while others may conclude that the tradeoffs simply aren’t worth it.

I suspect all of those decisions can be reasonable.

One of the things I’ve appreciated about writing and speaking about AI over the past year is that it continually reminds me that thoughtful people can arrive at different conclusions. Two therapists can look at the same technology, ask the same questions, and make different decisions based on their practice setting, client population, comfort level, and professional judgment.

That’s part of what makes these conversations both interesting and, at times, challenging.

If I’m being honest, I don’t know that we’re moving toward a future where every therapist uses AI in the same way. I suspect we’re moving toward a future where therapists become increasingly comfortable evaluating technology through the lens of their own values, responsibilities, and goals.

And perhaps that’s where this conversation ultimately led me.

I started by wanting to understand what it was like to build AI for therapy practices. Along the way, I heard about testing, crisis protocols, implementation challenges, privacy concerns, and the ongoing role of human oversight. But I also came away with the sense that many of the questions being asked by people building these tools aren’t all that different from the questions therapists are asking when deciding whether to use them.

What is this designed to do?

Where are its limitations?

How will it affect the people we serve?

And are we comfortable with the answers?

I don’t think discernment is a skill therapists will need only once as AI continues to evolve. If anything, I suspect it will become increasingly important over time.

The good news is that discernment is something therapists have been practicing all along.

It’s there every time we weigh competing priorities, make decisions with incomplete information, or consider the potential impact our choices may have on the people who trust us with their care.

Technology may change. The questions themselves may evolve. But the work of thoughtful decision-making has always been part of being a therapist.

And, at least for now, I suspect that isn’t changing anytime soon.

About Uriah Guilford

Uriah Guilford is a therapist, entrepreneur, and founder of Productive Therapist, where he supports mental health professionals with practice operations, intake coordination, and administrative services. He is also the founder of Simple Intake, an AI-supported intake and receptionist platform designed for therapy practices.

Learn more at Productive Therapist, the Productive Therapist Directory, and Simple Intake.

About the Interview

This article was informed by a conversation with therapist and entrepreneur Uriah Guilford, founder of Productive Therapist and Simple Intake. Our discussion explored AI implementation, privacy, human oversight, and the practical realities of building technology for therapy practices.

About the Author

Samantha Schalk, LMSW-C, LMSW-M, CAADC, CIMHP, BCP3

Samantha is a licensed mental health professional, private and group practice owner, and the founder of Guardian Clinical Essentials™.

She helps therapists and group practices understand how compliance, documentation, privacy, technology, and practice operations work together in real-world clinical settings. Her work focuses on turning complex requirements into practical systems, policies, workflows, and implementation strategies that providers can actually use.

Drawing from experience in both clinical practice and compliance consulting, Samantha specializes in helping mental health professionals build defensible, sustainable systems that support both quality care and regulatory compliance.

Learn more about Samantha and Guardian Clinical Essentials™.

Continue Exploring Guardian Clinical Essentials™

Looking for additional compliance, privacy, AI, documentation, and practice operations resources?

Explore the Guardian Clinical Essentials™ Resource Library for educational articles, implementation guidance, training opportunities, and practical resources designed specifically for mental health professionals.

GUARDIAN Clinical Essentials