Replace yourself. Then re-place yourself.

A lot of the conversation about AI in behavioral health comes back to one fear: that it’s coming for our jobs.

I understand where that comes from. I’m a social worker and a therapist, and I’ve worked in behavioral health for more than 15 years, starting back in college. I’ve worked on inpatient units and in children’s hospitals, provided trauma therapy, and spent years doing home-based work with families before supervising and building those programs in communities that didn’t have them. I’ve trained behavioral health professionals across Ohio, and now I lead a regional youth mobile crisis program.

Early on, doing individual therapy, I realized my clients needed much more than 50 minutes in a room with me. When a family trusts you enough to let you into their home, you see a whole different picture of what’s going on, and you can offer a whole different level of support. The work we do is built on human connection, and nobody wants a machine standing in for that.

My honest answer to that fear is that I hope AI replaces you in the parts of your day that were never the reason you got into this work. And then I want you to re-place yourself, on purpose, in the parts that were.

The replacing is the AI part

No therapist got into this work to write notes. No one chose this field so they could fight with insurance companies, worry about clawbacks, prepare for Medicaid audits, or write documentation that proves on paper that you’re actually doing good work. That part of the job exists for real reasons. Good documentation supports continuity of care and gives us room to reflect on the work, and I’m not saying it doesn’t matter. There has to be a process for how services get billed, and I understand the business side of it. But does it need to be you, writing out every word, for hours of your day? It doesn’t.

No therapist got into this work to write notes.

This isn’t a new problem for me. When I was a training manager, I built a workshop for clinicians on meeting productivity standards, organizing documentation, and Ohio Medicaid billing, and I wrote a documentation resource guide to go with it. We were teaching people how to survive the paperwork. AI is the first tool I’ve seen that can take some of it off their plate.

The same goes for everything around the clinical work: the spreadsheet sorting, the email writing, the loop closing, the to-do list that never ends. It even goes for the small stuff that quietly eats your brain, like how long you spend deciding whether that sentence in an email needs an exclamation point.

That’s what I mean by replacing yourself. You aren’t handing your job to AI. You’re handing it the tasks, so the person doing the job has more of herself left for the work only she can do.

The re-placing is the human part

Getting time back is only half of it, and it’s the half most AI advice stops at. You also have to decide where that time goes.

For some people it’s more time with clients, youth, and families. For others it’s building the system that would make everything run better, or finally writing the AI plan your organization doesn’t have yet. I know how easily that time disappears. I can finish something faster and immediately add another thing to the list. Innovation takes a lot of cognitive load. Most people I know in this field have plenty of ideas. What they don’t have is any mental space left at the end of the day to do anything with them.

One honest caution for leaders. In a productivity-driven agency, the hour a clinician saves on documentation can quietly become one more client on their caseload. If that’s all that happens, we haven’t given anyone their time back. We’ve just sped up the burnout. Re-placing has to be a decision someone makes on purpose, and in an organization, part of that decision belongs to leadership.

Being the only one who knows isn’t the goal

This one is a little uncomfortable to say out loud. Sometimes there’s ego in being the only person who knows something, because then everyone has to come to you. It can feel like being important.

It isn’t executive thinking. To be fair, ego isn’t the only reason someone becomes the go-to person. Sometimes there’s never been time to write anything down, the documentation that exists isn’t reliable, or no one is clear on who’s allowed to make the call. Those need fixing too. Whatever the reason, the fewer people who need to come to you for a sign-off or an answer, the more of your time goes to the work that moves things forward.

No one has to come to me, and I still know the information.

One idea I’ve been working through is pulling all the knowledge documents for a program I oversee into one place, so staff can get their questions answered without waiting on me. Answering a question and approving a decision aren’t the same thing, so anything that needs a sign-off would still come to a person. But for the questions I answer over and over, no one has to come to me, and I still know the information.

I went through my closet recently and pulled out stacks of paper training packets from over the years. All the notes I took, everything I learned, lives in my head or on that paper. It isn’t anywhere someone else could find it and use it. Most experienced people in our field have the same closet, or a stack of totes labeled “work stuff” that you still haven’t gotten around to organizing (and if you’re being real honest, they probably haven’t even been used since they went in there).

Context is what removes the burden

This is the part people miss when they try AI once and decide it takes longer than doing it themselves. Without relevant context, the answer can leave you doing most of the work again.

Say I ask ChatGPT for a budget-friendly grocery list for a family of six. It’ll give me one. But a good list depends on things it doesn’t know: allergies, how old the kids are, what they will and won’t eat, what my budget actually is, which stores I can get to. Either it guesses, or it asks me all of that, and next Sunday I answer every question again. At that point it would have been easier to walk into the store and shop.

Context is what removes the cognitive burden. Capture the stable parts once, reuse them, and update what changes. And you don’t have to sit down and type it out in some structured way. I talk mine out. Some of my best thinking lately has come from voice rambles where I just talk through what I know, because speaking and typing are different ways of processing. I figure things out as I talk.

You pulled that out of thin air. It’s actually more this than that.

Context is also what keeps your judgment in charge. AI can produce a note, a diagnosis summary, or a workflow that sounds amazing. Someone without your training would read it and nod. You’re the one who can look at it and say, “You pulled that out of thin air. It’s actually more this than that.” Keeping a human in the loop matters. That doesn’t mean spending hours fixing bad AI output. Good context cuts down on the fixing. How closely you review still depends on what’s at stake.

Count the whole job

An AI draft that shows up in seconds isn’t automatically a time saving. Count the whole job: the time it took to set it up, read it, fix it, and follow through. Then notice whether the work actually got lighter or just moved to someone else’s desk, like a supervisor who now has to check every AI-assisted note. If you keep fixing the same problems in the output, that’s a sign to look at your context, your process, or whether this is the right task for AI at all.

Be careful which tasks you hand off

There’s one more thing I’d weigh before replacing any task. Some of the work that feels like a burden is also how you got good.

I learned case conceptualization in MST without AI. We conceptualized every case deeply, and I had to take everything I knew about a family, distill my own thinking into a specific format, and then speak about each case clearly and concisely in supervision. That’s where a lot of my critical thinking came from. It’s also the skill I use now when I read an AI draft and know it’s wrong.

So before you hand something off, ask whether it’s only taking your time or whether it’s still building a skill you need. For an experienced clinician, AI drafting from your own thinking can save real time. For someone new to the field, doing that thinking themselves may be the training. Supervisors and agencies should make that call on purpose, because the judgment we’re counting on to check AI has to come from somewhere.

Start outside the client chart

When AI comes up in behavioral health, most of what I hear is about ethics. How do we protect our clients? How do we use it responsibly? Those questions matter, and the answers belong in agency policy all the way down to direct client care. But I don’t think they should be the first chapter for most people.

You don’t have to begin with clinical work. You can learn the most important AI habits on ordinary tasks that don’t involve client records or clinical decisions. Plan the grocery list. Draft the email you’ve been overthinking. Sort the spreadsheet. Pick something you know well enough to check the result. Share only the personal details the task actually needs, and never, ever put in a Social Security number.

That’s about your own information. Client information is a different conversation, and it belongs inside your agency’s approved tools and policies.

The same principles still apply. You still need your judgment, and context still does the heavy lifting. You’re just practicing where a mistake is easy to catch and cheap to fix.

Once you’ve built those habits, you’ll be in a much better position to bring AI into your work carefully, because you’ll know firsthand what it does well and where it makes things up.

Why this matters to me

My dad worked hard his entire life. Twelve-hour shifts, doubles, midnights straight into day shift, all while raising three kids on his own. He retired, and he died a month later. He wanted something different for me, and I think about that a lot.

I’ve worked hard too, and I can give my kids a lot of what they need. What I haven’t been able to give them enough of is me. Right now I’m mostly trading time from one place to another. That’s the real reason I started learning AI. I want my time back for the people I love.

Your reason might look different. Maybe you want more time to focus deeply on what you’re really passionate about, or room for more innovative work. Or maybe you just want to go outside and play with your kids without your mind screaming at you to get up and do something productive. Whatever your reason is, I want to help you get your time back too.

Where to start this week

The re-placing is up to you, and it’s worth naming where that time goes before the calendar fills it again.