I canceled my YMCA membership by sending a text to Instinct

A writer used Instinct, a personal AI agent accessed through iMessage, to cancel a household YMCA membership, with the agent ending family access on September 30 and stopping a $60 charge scheduled for October 1. The agent handled the YMCA website's dropped login connection, requested fresh verification codes, and confirmed the cancellation by email, after which the writer received a confirmation. The account notes Instinct competes with OpenClaw, an open-source agent that runs on a user's own machine, and Meta's Muse, a personal agent aimed at a mass audience.

At 8:13 a.m., I texted Instinct: “I want to cancel my YMCA membership.” I was on my phone. I didn’t have a laptop open, and I hadn’t decided that this was the moment I was going to deal with the YMCA. It had just crossed my mind, so I sent the text. Instinct is a personal AI agent that I use through iMessage. You can ask it to do things in much the same way you’d ask a person to help you. I’m trying it partly because I work in AI, and staying close to what’s happening means using the new tools on real problems. Demos are useful. Giving an agent access to your actual to-do list is where you find out what it can do. We are in a particularly interesting moment for personal agents. OpenClaw https://openclaw.ai/ showed what was possible if you connected an agent to your messages, accounts, and computer and let it take action. It’s open source and runs on your own machine; the flexibility is enormous, but you also have to set it up and maintain it. Meta’s Muse https://about.fb.com/news/2026/09/introducing-muse-personal-ai-agent/ is making a different bet: a personal agent for a mass audience that can use a browser, work across connected apps, and keep going when you close the app. Instinct’s bet, at least in my experience, is that an assistant should be as easy to reach as any other person in your messages. All three are trying to get at a problem that chatbots have mostly left with us. A chatbot can tell me how to cancel my YMCA membership. I still have to go cancel it. A personal agent is supposed to take the task and come back when it needs a decision. That sounds like a small distinction until you think about the things you keep putting off. The YMCA test Instinct asked which branch I belonged to. My phone number wasn’t on the account, so it asked for my email address. I forwarded a login code when it needed one. It found our household membership and came back with the information I would have needed before submitting anything. Cancelling would end access for the family on September 30 and stop a $60 charge scheduled for October 1. It said there was no outstanding balance, cancellation fee, add-on, or program registration to deal with. Then it asked if I wanted to proceed. I said yes and chose “not using it” as the reason. The YMCA website dropped the connection during login. Instinct told me nothing had been canceled and asked for a fresh code. It needed one more code before it got through. Then it reported that the membership was canceled and that there were no scheduled payments, and I received the confirmation email. By 11:30, it was texting me to ask what else I’d been meaning to get done. I can’t tell you the exact minute it completed the cancellation from the thread. I can tell you what I did: I sent a few short answers, forwarded the codes, and approved the cancellation. I never had to sit down at my desk. I think that last part is easy to miss if you’re evaluating this as a productivity tool. You might ask how long canceling the membership would have taken me on my own, subtract the time I spent texting, and conclude that the savings were modest. That’s the wrong math for this particular task. The reason the membership was still active wasn’t that I couldn’t find 15 minutes somewhere in my week. It’s that I needed 15 minutes at a computer , with the right login, the patience to work through whatever process the YMCA had built, and enough attention left over to check that it really canceled. There are plenty of those 15-minute tasks in a household. They accumulate precisely because each one feels too small to plan around and too irritating to do right now. With Instinct, I could start when I remembered. When it needed me, I could reply from my phone. The website failure was its problem to work through, with occasional help from me. That’s a much more interesting change than a few minutes saved on a stopwatch. Who owns the unfinished task? I spend a lot of time thinking about how AI changes work. One pattern keeps coming up: we celebrate the speed of the first step while ignoring everything that happens after it. Generating code is fast; reviewing and merging it can still take days. Drafting an email is fast; deciding what to say, sending it, and following up remain someone’s job. Personal agents have the same problem. The easy demo is an agent filling out a form. The actual test is what happens when the account is under a different email, the site drops its connection, a household membership is involved, and a charge is about to hit. Does the agent keep track of where it is? Does it tell me when nothing has happened yet? Does it ask before making a consequential choice? At the end, can I take the task off my list? Instinct did well on those questions in this case. When the site failed, it didn’t claim success. Before submitting, it gave me the end date and the charge I would avoid, then asked for approval. Afterward, I had a confirmation email. The point wasn’t that I never had to touch the task. I was still the person who could authorize canceling my family’s membership. The point was that I didn’t have to own every step between the initial request and the result. Of course, that kind of help requires access. An agent can’t navigate my accounts without some way into my accounts. It can’t learn my preferences without knowing things about me. OpenClaw, Muse, and Instinct make different choices about who sets up that access and where the agent runs, but anyone using a personal agent has to decide what they’re comfortable handing over. The more useful these tools become, the less theoretical that decision feels. I’m not ready to declare that my entire life should run through an AI assistant because it canceled one membership. But I am paying attention to the kind of task it handled. The value of a personal agent won’t be measured by how impressively it talks about the things on my to-do list. It will be measured by how many of those things I no longer have to carry around in my head. The YMCA cancellation is now off my list. I didn’t have to be at my desk to make it happen. That’s a pretty good first test.