I’ve spent a little over a week using Dots from ChatGPT as an AI assistant for my business, and I named mine Dottie. Some of what we’ve worked on together has been really useful. There have also been corrections, and a few moments where I’ve had to stop and ask, wait, where is that? In this episode of Run Your Private Practice with AI, I walk through the ordinary, nonclinical tasks I’ve handed off so far, from my inbox to a resource we’re building to a reference request that didn’t go as planned. If you’ve been curious about using Dots from ChatGPT or another AI agent in your practice, this is an honest look at what the setup takes and where you still need to stay involved. AI Tips You Don’t Want to Miss: What “clean up my inbox” leaves open, and what I had to spell out. Why I turned a confusing guide into a file an AI agent can walk you through. The bounced email that showed me where a handed-off task can stall. Where I still have to step in, from accounting questions to how my writing sounds. One small, nonclinical place to start testing an agent this week. Mentioned in this episode: Thera AI Hub waitlist: https://theraaihub.netlify.app/#tah-waitlist ChatGPT dots overview and current access information: https://learn.chatgpt.com/docs/dots ___________________________ Connect with me: Instagram: @thetravelingtherapist_kym Website: AIfortherapists.com Clinical AI Club and Thera AI Hub are merging into one larger program. Join the waitlist to be the first to hear about it: www.theraaihub.netlify.app The Show is Sponsored by: Berries: Say goodbye to the burden of mental health notes with automated note and treatment plan creation! www.heyberries.com/therapists Auralink Pro: Give clients guided support between sessions and get a clinical summary before they’re back on your couch! Try it free for 30 days (no credit card needed) plus 30% off your first 12 months. www.auralink.com/kym
Audio and video play straight from the show's own hosting. PodSubs stores and serves no media.
Transcript
From the show. Tap a time to play from there.
SPEAKER_00Welcome to Run Your Private Practice with AI, the podcast where innovation meets inspiration. I created this podcast to help therapists like you harness the power of AI, transforming the way you manage your private practice and care for your clients. Join us as we explore the dynamic world of AI, unlocking new possibilities for efficiency, creativity, and personal growth as therapists in private practice.
SPEAKER_01And then I feel like we're missing important things. So today I want to get into that. But before we do, welcome back to another episode of Run Your Private Practice with AI. This is the podcast where I try to keep AI simple. I try to keep you up to date with what's happening with AI and also keep you informed about tools you can use to help you run your private practice and your business just in general. We were working on emails. There were people I needed to get back to, drafts we'd talked about, and I was trying to remember where everything was.
SPEAKER_01I wanted to be able to look somewhere and know, okay, here's what still needs me. So that's probably a useful place to start an episode about an AI tool that I'm really impressed with. Because there have been corrections. There have been moments where I've said, wait, where is that? And there have also been some really useful things we've worked through together. I've been using ChatGPT Dots for over a week, and I named mine Dottie. So when I say Dottie, that's what I'm talking about. My AI assistant inside ChatGPT.
SPEAKER_01Before this, I was using Instinct, which I think is very good. And so far, for the way I'm using it, I think I like dots even better. That's my experience. I haven't done some formal comparison of every feature, and I'm not saying everybody's going to prefer the same thing, but I wanted to walk you through what we've actually been doing, because I think hearing a few ordinary examples makes this easier to understand than hearing that an AI agent can help you run your business. Okay, help me do what? If you're already using AI to write something or help you think through an idea, some of this will feel familiar.
SPEAKER_01With dots, you can also give it an ongoing responsibility. Depending on the access you've connected and the permissions you've given it, it can work between conversations and come back with a result or something you need to decide. That does require setup, and it doesn't mean it can do every action in every app or that you should give it access to everything you have. I'm going to keep the examples here on the business side. I'm not talking about putting client information into a general AI assistant. That needs its own privacy and security review.
SPEAKER_01Okay, so let's start with the email situation. I told Dottie I wanted my inbox to have the things I needed to reply to or know about, which sounds clear until you start looking at the emails. A newsletter I don't need can go, but an email that looks like a newsletter might have a resource I paid for, or an opportunity that actually fits my business. A cancellation needs to go to the person on my team who handles those. It shouldn't disappear just because I personally don't need to answer it. We had to get specific about those differences.
SPEAKER_01And one of the things I wanted was for replay links from paid memberships to go into my replay spreadsheet. Because if I paid for the training, I want to be able to find the recording later. I don't want the email gone and then have to remember which membership it came from. So the instruction includes saving that link and checking that it actually saved before filing the email. If it can't do that, the email needs to stay where we can find it. That's the kind of detail I think matters when you're handing something off. Clean up my inbox leaves a lot open to interpretation.
SPEAKER_01I had to explain what I wanted protected, and I kept changing or clarifying things as we went. I might say, yes, those notices can go, but leave that one. Or put this type of request in this folder for my VA. There were also actions it couldn't complete, so it had to stop and tell me instead of acting like the cleanup was finished. I'm not going to tell you my inbox is now perfect. You heard where I started. I was asking about important things I couldn't find. That part matters too. I still need a way to see what's waiting for my review, and I need to know whether a reply is sitting in our conversation or actually saved as a draft in Outlook.
SPEAKER_01Those are different places. Saying, we wrote that email doesn't tell me where to go to send it. So if you try something like this, make that part of the instruction. Where should the draft go? What should stay visible? And how should the assistant tell you something couldn't be completed? Another example is a freebie we've been working on. I wanted a helpful resource for another business owner's audience. We were building a follow-through workflow, something that could help a practice owner deal with the business tasks that need another email or a decision before they can move forward.
SPEAKER_01And when I looked at the guide, I told Dottie it was confusing. I thought we were giving people more work to do. Because if I hand you a whole process and you have to read it and organize your tasks and figure out how to run it, are you going to use it? You already have the tasks. Now you also have a guide to manage. What I wanted was a file you could give to an AI agent so the agent would understand the workflow and walk you through it. You could come in with a messy list. You wouldn't have to get everything perfectly organized first. The agent could help you figure out what's waiting on somebody else and what needs you to make a decision.
SPEAKER_01It could help draft that follow-up email while you're right there. We revised it and I tested it. Then I asked for simulated conversations with the kind of practice owner who would actually receive it. Those were fictional test users, so that isn't the same as having real customers use it or proving it works on every AI platform, but it gave us something useful to inspect. What happens if the person says, I'm not techie, go slow? Does the next response actually get shorter? What happens if they come back a week later? Does the AI ask for an update, or does it act like it knows what happened when it doesn't?
SPEAKER_01And here's a detail I really wanted us to catch. If I sent a follow-up email, that doesn't necessarily mean the task is done. I might still be waiting for a reply. We needed the workflow to keep that distinction. We also talked about reminders. Rather than assuming everyone wants a recurring reminder set up, why don't we just ask whether they want one? These are small things when you're looking at a file. They're pretty important when someone's actually using it. We're still working through that resource. I'm not presenting it as a finished, tested everywhere product, but this gives you an idea of how I'm using Dottie to develop something.
SPEAKER_01I can bring back the part that doesn't work and say, okay, let's fix this part before we put it in front of people. And yes, I also asked for the PDF to have a white background because I don't like beige backgrounds on things. That's apparently part of the instructions now, too. Here's another example that doesn't involve creating content at all. I needed written recommendations for someone applying to a vetted level of my VA directory. I asked Dottie to email the references and explain what we were asking for. That was something I specifically asked it to send.
SPEAKER_01With other emails, I've wanted a draft so I can review it first. Those permissions shouldn't get mixed together. One of the reference emails bounced. The address we'd been given wasn't right. So the next step was to ask the applicant for the corrected address, then send the request to the corrected address. We still had references we were waiting on. I like that example because it's so ordinary. You send an email and there's still a little bit of work after that. Did it go through? Do we have a response? Is there something I need to do before this can move forward?
SPEAKER_01For a therapist running a business, there are plenty of non-clinical versions of that. Maybe you're waiting on a website designer to change a page, or you need a vendor to answer a question before you renew something. You could have an assistant help you keep track of that, with clear instructions about what it's allowed to send and what you want to review. I wouldn't start with the most sensitive thing in your business. Start somewhere you can easily check the result. We also worked on questions from my accountant about transactions. Dottie helped research what some of the charges might be and helped me prepare a reply.
SPEAKER_01I had to supply context and correct how one transaction was being understood. That context wasn't something it could just know from a description. So I still needed to be part of it, and my accountant still needed to make the accounting decisions. I could get help gathering the information without treating the AI's first explanation as the answer. I'm also using it to help with AI research for content. I've asked it to read relevant newsletters and check the original sources behind the stories, then help me turn worthwhile topics into drafts for my audience.
SPEAKER_01And I care about the difference between an announcement and something people can actually use. If a company announces a feature, that doesn't automatically mean every listener can open their account and find it today. That applies to dots too. Access is rolling out, so check the current official information for your account. I'll include that link with this episode rather than give you a list of plans that could be out of date when you're listening. The writing still needs my input. I've told Dottie that I use contractions when I speak. I've asked for things to sound more like me.
SPEAKER_01And with some AI video experiments, I've said the face didn't look like me, the delivery was too excited, and then the pacing was too fast or too slow. We're not going to call something finished just because a tool produced a video. I need to be comfortable putting my name on it. So if you're listening and thinking, okay, where would I even start with an agent? I pick one recurring business task that's easy to review. Maybe that's a short list of things you're waiting on. You can keep it non-clinical. Something like, the web designer said they'd send a revised page.
SPEAKER_01I'm waiting for the invoice from this vendor. I need to decide whether I'm renewing this subscription. Tell the assistant what you want it to do with that list. You might say, help me review this when I check in. Ask me what's changed. Draft a follow-up if we need one, but don't send it. Keep anything unresolved on the list. Then run through a real example and read what it gives you. If it assumes something happened, correct that. If it asks you a whole questionnaire when you wanted one question, tell it. You don't have to accept the first version because it sounds organized.
SPEAKER_01And if you want it to work on a schedule, check that scheduling is available and actually set up. A conversation about checking something every Monday isn't proof that a reminder or background task exists. I think that's part of learning to use these tools well. You get clearer about what you want and you check what happened. Over time, you can decide whether you're comfortable handing over more of a particular workflow. And next month, in November, this is what we're going to focus on inside Thera AI Hub. I'm going to compare dots, Instinct, Q from Manus, and Duet from Claude so you can see how these different agents work and start figuring out which one makes sense for you.
SPEAKER_01I'm also planning to give members five workflows they can upload to an agent, like the follow-through workflow I talked about earlier. We'll look at how to get those set up and where the agent supports it and has the right access, have them running in the background. I want you to have something you can actually try in your business while you're learning how this works. Thera AI Hub is on a wait list right now, and I'm aiming to open it at the end of November. If you'd like to join us, get on the wait list using the link below this episode. I've been pretty impressed by what we've been able to work on in a little over a week.
SPEAKER_01I also have things I'm still correcting and figuring out. Both of those things can be true. And I'm curious where this goes as more people start trying agents in the ordinary parts of running a practice. The email you need to answer, the resource you want to find again, the project that's waiting on somebody to get back to you. So that's my suggestion for this week. Choose one of those small, non-clinical tasks and test an agent on it. Give it a clear boundary, then look at what it actually did. See whether it helps with something you genuinely need help with.
SPEAKER_01And if you hear me mentioning Dottie in another episode, now you know who I'm talking about.
There’s always more a client needs to say than fits in one session, and that gap is a big part of why I’ve been spending so much time with Auralink Pro. It’s built for therapists; it isn’t trying to do therapy, and it’s...
Keeping up with AI as a therapist can feel like a part-time job, which is a big reason I rebuilt everything from the ground up into the new Thera AI Hub. The tools change every week, and I wanted members to stop...
Most of what wears you down in private practice happens around the appointments, and a time study is one of the simplest ways for therapists to see where those hours actually go. Those small business tasks add up...
The question I get asked most about AI notes is how to tell clients you use AI. More specifically, people want to know what to say out loud to someone who has trusted them with the hardest parts of their life. In this...
When someone asks ChatGPT to find a therapist who understands burnout, offers evening appointments, and works online, AEO optimization plays a big part in whether your practice makes that list. Their friend could type...